autogluon.multimodal 1.1.2b20241106__py3-none-any.whl → 1.1.2b20241108__py3-none-any.whl

autogluon/multimodal/configs/model/default.yaml

@@ -115,6 +115,7 @@ model:
     max_img_num_per_col: 1
     output_bbox_format: "xyxy"  # now support xyxy or xywh, for bbox format details see https://keras.io/api/keras_cv/bounding_box/formats/
     frozen_layers: null
+    coco_root: null
 
   mmocr_text_detection:
     checkpoint_name: "TextSnake"

autogluon/multimodal/configs/pretrain/detection/dino/dino-4scale_r50_8xb2-12e_coco.py

@@ -84,7 +84,7 @@ model = dict(
 # train_pipeline, NOTE the img_scale and the Pad's size_divisor is different
 # from the default setting in mmdet.
 train_pipeline = [
-    dict(type="LoadImageFromFile", backend_args={{_base_.backend_args}}),
+    dict(type="LoadImageFromFile", backend_args={{_base_.backend_args}}),  # nosec
     dict(type="LoadAnnotations", with_bbox=True),
     dict(type="RandomFlip", prob=0.5),
     dict(

autogluon/multimodal/learners/object_detection.py

@@ -15,12 +15,13 @@ from ..utils import (
     check_if_packages_installed,
     cocoeval,
     convert_pred_to_xywh,
+    convert_result_df,
     create_fusion_model,
     extract_from_output,
     from_coco_or_voc,
     get_detection_classes,
     object_detection_data_to_df,
-    save_result_df,
+    save_result_coco_format,
     setup_save_path,
     split_train_tuning_data,
 )
@@ -39,8 +40,9 @@ class ObjectDetectionLearner(BaseLearner):
         hyperparameters: Optional[dict] = None,
         path: Optional[str] = None,
         verbosity: Optional[int] = 2,
-        num_classes: Optional[int] = None,  # TODO: can we infer this from train/predict data?
-        classes: Optional[list] = None,  # TODO: can we infer this from train/predict data?
+        num_classes: Optional[int] = None,
+        classes: Optional[list] = None,
+        category_ids: Optional[list] = None,
         warn_if_exist: Optional[bool] = True,
         enable_progress_bar: Optional[bool] = None,
         pretrained: Optional[bool] = True,
@@ -74,14 +76,17 @@ class ObjectDetectionLearner(BaseLearner):
         )
         check_if_packages_installed(problem_type=self._problem_type)
 
+        self._config = self.get_config_per_run(config=self._config, hyperparameters=hyperparameters)
+
         self._output_shape = num_classes
         self._classes = classes
+        self._category_ids = category_ids
         self._sample_data_path = sample_data_path
 
         # TODO: merge object detection and open vocabulary object detection
         self._label_column = "label"
         if self._sample_data_path is not None:
-            self._classes = get_detection_classes(self._sample_data_path)
+            self._classes, self._category_ids = get_detection_classes(self._sample_data_path)
             self._output_shape = len(self._classes)
 
         # TODO: merge _detection_anno_train and detection_anno_train?
@@ -97,15 +102,32 @@ class ObjectDetectionLearner(BaseLearner):
         """
         Return the classes of object detection.
         """
-        return self._model.model.CLASSES
+        if self._model.model.CLASSES is not None and self._classes is not None:
+            assert self._classes == self._model.model.CLASSES, f"{self._classes}\n{self._model.model.CLASSES}"
+        return self._classes if self._classes is not None else self._model.model.CLASSES
+
+    @property
+    def category_ids(self):
+        """
+        Return the classes of object detection.
+        """
+        return self._category_ids
 
     def setup_detection_train_tuning_data(self, max_num_tuning_data, seed, train_data, tuning_data):
         if isinstance(train_data, str):
             self._detection_anno_train = train_data
-            train_data = from_coco_or_voc(train_data, "train")  # TODO: Refactor to use convert_data_to_df
+            train_data = from_coco_or_voc(
+                train_data,
+                "train",
+                coco_root=self._config.model.mmdet_image.coco_root,
+            )  # TODO: Refactor to use convert_data_to_df
             if tuning_data is not None:
                 self.detection_anno_train = tuning_data
-                tuning_data = from_coco_or_voc(tuning_data, "val")  # TODO: Refactor to use convert_data_to_df
+                tuning_data = from_coco_or_voc(
+                    tuning_data,
+                    "val",
+                    coco_root=self._config.model.mmdet_image.coco_root,
+                )  # TODO: Refactor to use convert_data_to_df
                 if max_num_tuning_data is not None:
                     if len(tuning_data) > max_num_tuning_data:
                         tuning_data = tuning_data.sample(
@@ -114,10 +136,16 @@ class ObjectDetectionLearner(BaseLearner):
         elif isinstance(train_data, pd.DataFrame):
             self._detection_anno_train = None
             # sanity check dataframe columns
-            train_data = object_detection_data_to_df(train_data)
+            train_data = object_detection_data_to_df(
+                train_data,
+                coco_root=self._config.model.mmdet_image.coco_root,
+            )
             if tuning_data is not None:
                 self.detection_anno_train = tuning_data
-                tuning_data = object_detection_data_to_df(tuning_data)
+                tuning_data = object_detection_data_to_df(
+                    tuning_data,
+                    coco_root=self._config.model.mmdet_image.coco_root,
+                )
                 if max_num_tuning_data is not None:
                     if len(tuning_data) > max_num_tuning_data:
                         tuning_data = tuning_data.sample(
@@ -556,7 +584,9 @@ class ObjectDetectionLearner(BaseLearner):
         if isinstance(anno_file_or_df, str):
             anno_file = anno_file_or_df
             data = from_coco_or_voc(
-                anno_file, "test"
+                anno_file,
+                "test",
+                coco_root=self._config.model.mmdet_image.coco_root,
             )  # TODO: maybe remove default splits hardcoding (only used in VOC)
             if os.path.isdir(anno_file):
                 eval_tool = "torchmetrics"  # we can only use torchmetrics for VOC format evaluation.
@@ -636,7 +666,10 @@ class ObjectDetectionLearner(BaseLearner):
                 eval_tool=eval_tool,
             )
         else:
-            data = object_detection_data_to_df(data)
+            data = object_detection_data_to_df(
+                data,
+                coco_root=self._config.model.mmdet_image.coco_root,
+            )
             return self.evaluate_coco(
                 anno_file_or_df=data,
                 metrics=metrics,
@@ -648,6 +681,7 @@ class ObjectDetectionLearner(BaseLearner):
         self,
         data: Union[pd.DataFrame, dict, list, str],
         as_pandas: Optional[bool] = None,
+        as_coco: Optional[bool] = True,
         realtime: Optional[bool] = False,
         save_results: Optional[bool] = None,
         **kwargs,
@@ -662,6 +696,8 @@ class ObjectDetectionLearner(BaseLearner):
             follow same format (except for the `label` column).
         as_pandas
             Whether to return the output as a pandas DataFrame(Series) (True) or numpy array (False).
+        as_coco
+            Whether to save the output as a COCO json file (True) or pandas DataFrame (False).
         realtime
             Whether to do realtime inference, which is efficient for small data (default False).
             If provided None, we would infer it on based on the data modalities
@@ -674,46 +710,65 @@ class ObjectDetectionLearner(BaseLearner):
         Array of predictions, one corresponding to each row in given dataset.
         """
         self.ensure_predict_ready()
+        if as_pandas is None and isinstance(data, pd.DataFrame):
+            as_pandas = True
+
         ret_type = BBOX
+
+        # only supports coco/voc format for OBJECT_DETECTION
         if self._problem_type == OBJECT_DETECTION:
-            data = object_detection_data_to_df(data)
-            if self._label_column not in data:
+            data_path = data
+            data_df = object_detection_data_to_df(
+                data_path,
+                coco_root=self._config.model.mmdet_image.coco_root,
+            )
+            if self._label_column not in data_df:
                 self._label_column = None
 
         outputs = self.predict_per_run(
-            data=data,
+            data=data_df,
             realtime=realtime,
             requires_label=False,
         )
         pred = extract_from_output(outputs=outputs, ret_type=ret_type)
-        if self._problem_type == OBJECT_DETECTION:
-            if self._model.output_bbox_format == XYWH:
-                pred = convert_pred_to_xywh(pred)
 
-        if save_results and self._problem_type == OBJECT_DETECTION:
+        self._save_path = setup_save_path(
+            old_save_path=self._save_path,
+            warn_if_exist=False,
+        )
+        result_path = os.path.join(self._save_path, "result.txt")
+
+        pred_df = convert_result_df(
+            pred=convert_pred_to_xywh(pred) if self._model.output_bbox_format == XYWH else pred,
+            data=data_df,
+            detection_classes=self.classes,
+            result_path=result_path,
+        )
+
+        if save_results:
             self._save_path = setup_save_path(
                 old_save_path=self._save_path,
                 warn_if_exist=False,
             )
-            save_result_df(
-                pred=pred,
-                data=data,
-                detection_classes=self._model.model.CLASSES,
-                result_path=os.path.join(self._save_path, "result.txt"),
-            )
-
-        if (as_pandas is None and isinstance(data, pd.DataFrame)) or as_pandas is True:
-            # TODO: add prediction output in COCO format if as_pandas is False
-            # TODO: calling save_result_df to convert data to dataframe is not a good logic
-            # TODO: consider combining this with the above saving logic or using a different function.
-            pred = save_result_df(
-                pred=pred,
-                data=data,
-                detection_classes=self._model.model.CLASSES,
-                result_path=None,
-            )
+            if as_coco:
+                result_path = os.path.join(self._save_path, "result.json")
+                save_result_coco_format(
+                    data_path=data_path,
+                    pred=pred,
+                    category_ids=self.category_ids,
+                    result_path=result_path,
+                    coco_root=self._config.model.mmdet_image.coco_root,
+                )
+            else:
+                pred_df.to_csv(result_path, index=False)
+            logger.info(f"Saved detection results {'as coco' if as_coco else 'as dataframe'} to {result_path}")
 
-        return pred
+        if as_pandas:
+            return pred_df
+        else:
+            if self._model.output_bbox_format == XYWH:
+                pred = convert_pred_to_xywh(pred)
+            return pred
 
     def predict_proba(
         self,

autogluon/multimodal/predictor.py

@@ -605,6 +605,7 @@ class MultiModalPredictor:
         as_pandas: Optional[bool] = None,
         realtime: Optional[bool] = False,
         save_results: Optional[bool] = None,
+        **kwargs,
     ):
         """
         Predict the label column values for new data.
@@ -617,8 +618,8 @@ class MultiModalPredictor:
         candidate_data
             The candidate data from which to search the query data's matches.
         id_mappings
-             Id-to-content mappings. The contents can be text, image, etc.
-             This is used when data contain the query/response identifiers instead of their contents.
+            Id-to-content mappings. The contents can be text, image, etc.
+            This is used when data contain the query/response identifiers instead of their contents.
         as_pandas
             Whether to return the output as a pandas DataFrame(Series) (True) or numpy array (False).
         realtime
@@ -627,10 +628,14 @@ class MultiModalPredictor:
             and sample number.
         save_results
             Whether to save the prediction results (only works for detection now)
+        **kwargs
+            Additional keyword arguments to pass to the underlying learner's predict method.
+            For example, `as_coco` for object detection tasks.
 
         Returns
         -------
         Array of predictions, one corresponding to each row in given dataset.
+        Format depends on the specific learner and provided arguments.
         """
         return self._learner.predict(
             data=data,
@@ -639,6 +644,7 @@ class MultiModalPredictor:
             realtime=realtime,
             save_results=save_results,
             id_mappings=id_mappings,
+            **kwargs,
         )
 
     def predict_proba(

autogluon/multimodal/utils/__init__.py

@@ -82,6 +82,7 @@ from .object_detection import (
     bbox_xyxy_to_xywh,
     cocoeval,
     convert_pred_to_xywh,
+    convert_result_df,
     from_coco,
     from_coco_or_voc,
     from_dict,
@@ -90,7 +91,6 @@ from .object_detection import (
     object_detection_data_to_df,
     object_detection_df_to_coco,
     save_result_coco_format,
-    save_result_df,
     save_result_voc_format,
     visualize_detection,
 )

autogluon/multimodal/utils/hpo.py

@@ -50,9 +50,9 @@ def hpo_trial(sampled_hyperparameters, learner, checkpoint_dir=None, **_fit_args
     resources = context.get_trial_resources().required_resources
     num_cpus = int(resources.get("CPU"))
 
-    _fit_args["hyperparameters"] = (
-        sampled_hyperparameters  # The original hyperparameters is the search space, replace it with the hyperparameters sampled
-    )
+    # The original hyperparameters is the search space, replace it with the hyperparameters sampled
+    _fit_args["hyperparameters"] = sampled_hyperparameters
+
     _fit_args["save_path"] = context.get_trial_dir()  # We want to save each trial to a separate directory
     logger.debug(f"hpo trial save_path: {_fit_args['save_path']}")
     if checkpoint_dir is not None:
@@ -62,7 +62,15 @@ def hpo_trial(sampled_hyperparameters, learner, checkpoint_dir=None, **_fit_args
         learner.fit_per_run(**_fit_args)
 
 
-def build_final_learner(learner, best_trial_path, save_path, last_ckpt_path, is_matching, standalone, clean_ckpts):
+def build_final_learner(
+    learner,
+    best_trial_path,
+    save_path,
+    last_ckpt_path,
+    is_matching,
+    standalone,
+    clean_ckpts,
+):
     """
     Build the final learner after HPO is finished.
 
@@ -167,12 +175,13 @@ def hyperparameter_tune(hyperparameter_tune_kwargs, resources, is_matching=False
     mode = _fit_args.get("learner")._minmax_mode
     save_path = _fit_args.get("save_path")
     time_budget_s = _fit_args.get("max_time")
+    num_to_keep = hyperparameter_tune_kwargs.pop("num_to_keep", 3)
     if time_budget_s is not None:
         time_budget_s *= 0.95  # give some buffer time to ray
     try:
         run_config_kwargs = {
             "checkpoint_config": CheckpointConfig(
-                num_to_keep=3,
+                num_to_keep=num_to_keep,
                 checkpoint_score_attribute=metric,
             ),
         }

autogluon/multimodal/utils/object_detection.py

@@ -9,6 +9,7 @@ import defusedxml.ElementTree as ET
 import numpy as np
 import pandas as pd
 import PIL
+import torch
 from torchmetrics.detection.mean_ap import MeanAveragePrecision
 
 from ..constants import (
@@ -58,24 +59,43 @@ def _get_image_info(image_path: str):
 
 def get_df_unique_classes(data: pd.DataFrame):
     """
-    Get the unique classes in the dataframe for object detection
+    Get the unique classes and their category IDs from the dataframe for object detection.
+
     Parameters
     ----------
-    data
-        pd.DataFrame holding the data for object detection
+    data : pd.DataFrame
+        DataFrame holding the data for object detection. Each row should contain a 'rois'
+        column with detection boxes and class labels.
+
     Returns
     -------
-        list of unique classes
+    tuple
+        A tuple containing (class_names, category_ids) where:
+        - class_names: list of unique class name strings
+        - category_ids: dict mapping class names to their numeric IDs
     """
     unique_classes = {}
+
+    # Iterate through all rows in the dataframe
     for idx in range(data.shape[0]):
         row = data.iloc[idx]
         rois = row["rois"]
+
+        # Process each ROI in the current row
         for roi in rois:
-            _, _, _, _, class_label = roi
+            # Unpack ROI values (assuming last element is class label)
+            *_, class_label = roi
+
+            # Add new classes to the dictionary with auto-incrementing IDs
             if class_label not in unique_classes:
-                unique_classes[class_label] = len(unique_classes)
-    return list(unique_classes.keys())
+                # Start IDs from 1, as 0 is often reserved for background
+                unique_classes[class_label] = len(unique_classes) + 1
+
+    # Create the output lists/dicts
+    class_names = list(unique_classes.keys())
+    category_ids = unique_classes
+
+    return class_names, category_ids
 
 
 def object_detection_df_to_coco(data: pd.DataFrame, save_path: Optional[str] = None):
@@ -145,7 +165,9 @@ def object_detection_df_to_coco(data: pd.DataFrame, save_path: Optional[str] = N
     return output_json_dict
 
 
-def object_detection_data_to_df(data: Union[pd.DataFrame, dict, list, str]) -> pd.DataFrame:
+def object_detection_data_to_df(
+    data: Union[pd.DataFrame, dict, list, str], coco_root: Optional[str] = None
+) -> pd.DataFrame:
     """
     Construct a dataframe from a data dictionary, json file path (for COCO), folder path (for VOC),
     image path (for single image), list of image paths (for multiple images)
@@ -163,7 +185,7 @@ def object_detection_data_to_df(data: Union[pd.DataFrame, dict, list, str]) -> p
         return from_list(data)
     if isinstance(data, str):
         if os.path.isdir(data) or data.endswith(".json"):
-            return from_coco_or_voc(data)
+            return from_coco_or_voc(data, coco_root=coco_root)
         return from_str(data)
     if isinstance(data, pd.DataFrame):
         sanity_check_dataframe(data)
@@ -303,7 +325,7 @@ def from_voc(
     rpath = Path(root).expanduser()
     img_list = []
 
-    class_names = get_detection_classes(root)
+    class_names, _ = get_detection_classes(root)
 
     NAME_TO_IDX = dict(zip(class_names, range(len(class_names))))
     name_to_index = lambda name: NAME_TO_IDX[name]
@@ -700,7 +722,7 @@ def _check_load_coco_bbox(
 
 def from_coco(
     anno_file: Optional[str],
-    root: Optional[str] = None,
+    coco_root: Optional[str] = None,
     min_object_area: Optional[Union[int, float]] = 0,
     use_crowd: Optional[bool] = False,
 ):
@@ -718,7 +740,7 @@ def from_coco(
     ----------
     anno_file
         The path to the annotation file.
-    root
+    coco_root
         Root of the COCO folder. The default relative root folder (if set to `None`) is `anno_file/../`.
     min_object_area
         Minimum object area to consider.
@@ -740,16 +762,19 @@ def from_coco(
     coco = COCO(anno_file)
 
     # get data root
-    if isinstance(root, Path):
-        root = str(root.expanduser().resolve())
-    elif isinstance(root, str):
-        root = os.path.abspath(os.path.expanduser(root))
-    elif root is None:
+    if isinstance(coco_root, Path):
+        coco_root = str(coco_root.expanduser().resolve())
+    elif isinstance(coco_root, str):
+        coco_root = os.path.abspath(os.path.expanduser(coco_root))
+    elif coco_root is None:
         # try to use the default coco structure
-        root = os.path.join(os.path.dirname(anno_file), "..")
-        logger.info(f"Using default root folder: {root}. Specify `root=...` if you feel it is wrong...")
+        coco_root = os.path.join(os.path.dirname(anno_file), "..")
+        logger.info(
+            f"Using default root folder: {coco_root}. "
+            "Specify `model.mmdet_image.coco_root=...` in hyperparameters if you think it is wrong."
+        )
     else:
-        raise ValueError("Unable to parse root: {}".format(root))
+        raise ValueError("Unable to parse coco_root: {}".format(coco_root))
 
     # support prediction using data with no annotations
     # note that data with annotation can be used for prediction without any changes
@@ -764,9 +789,9 @@ def from_coco(
     for entry in coco.loadImgs(image_ids):
         if "coco_url" in entry:
             dirname, filename = entry["coco_url"].split("/")[-2:]
-            abs_path = os.path.join(root, dirname, filename)
+            abs_path = os.path.join(coco_root, dirname, filename)
         else:
-            abs_path = os.path.join(root, entry["file_name"])
+            abs_path = os.path.join(coco_root, entry["file_name"])
         if not os.path.exists(abs_path):
             logger.warning(f"File skipped since not exists: {abs_path}.")
             continue
@@ -809,7 +834,7 @@ def get_image_filename(path: str):
 class COCODataset:
     # The class that load/save COCO data format.
     # TODO: refactor data loading into here
-    def __init__(self, anno_file: str):
+    def __init__(self, anno_file: str, category_ids: Optional[List] = None):
         """
         Parameters
         ----------
@@ -828,7 +853,13 @@ class COCODataset:
             img_id_list.append(int(img["id"]))
         self.image_filename_to_id = dict(zip(img_filename_list, img_id_list))
 
-        self.category_ids = [cat["id"] for cat in d["categories"]]
+        if category_ids is None:
+            if "categories" in d:
+                self.category_ids = [cat["id"] for cat in d["categories"]]
+            else:
+                self.category_ids = range(9999)  # TODO: remove hardcoding here
+        else:
+            self.category_ids = category_ids
 
     def get_image_id_from_path(self, image_path: str):
         """
@@ -1146,7 +1177,7 @@ def process_voc_annotations(
         f.writelines("\n".join(xml_file_names))
 
 
-def from_coco_or_voc(file_path: str, splits: Optional[Union[str]] = None):
+def from_coco_or_voc(file_path: str, splits: Optional[Union[str]] = None, coco_root: Optional[str] = None):
     """
     Convert the data from coco or voc format to pandas Dataframe.
 
@@ -1167,54 +1198,68 @@ def from_coco_or_voc(file_path: str, splits: Optional[Union[str]] = None):
         # VOC use dir as input
         return from_voc(root=file_path, splits=splits)
     else:
-        return from_coco(file_path)
+        return from_coco(file_path, coco_root=coco_root)
 
 
 def get_coco_format_classes(sample_data_path: str):
     """
-    The all class names for COCO format data.
+    Get class names and category IDs for COCO format data.
 
     Parameters
     ----------
-    sample_data_path
+    sample_data_path : str
         The path to COCO format json annotation file. Could be any split, e.g. train/val/test/....
 
     Returns
     -------
-        All the class names.
+    tuple
+        A tuple containing (class_names, category_ids) where:
+        - class_names: list of class name strings
+        - category_ids: dict mapping class names to their COCO category IDs
     """
     try:
         with open(sample_data_path, "r") as f:
             annotation = json.load(f)
     except:
         raise ValueError(f"Failed to load json from provided json file: {sample_data_path}.")
-    return [cat["name"] for cat in annotation["categories"]]
+
+    # Extract both names and IDs from categories
+    class_names = [cat["name"] for cat in annotation["categories"]]
+
+    # Create mapping of names to their original COCO category IDs
+    category_ids = [cat["id"] for cat in annotation["categories"]]
+
+    return class_names, category_ids
 
 
 def get_voc_format_classes(root: str):
     """
-    The all class names for VOC format data.
+    Get class names and category IDs for VOC format data.
 
     Parameters
     ----------
-    root
+    root : str
         The path to the root directory of VOC data.
 
     Returns
     -------
-        All the class names.
+    tuple
+        A tuple containing (class_names, category_ids) where:
+        - class_names: list of class name strings
+        - category_ids: dict mapping class names to their numeric IDs
     """
     if is_url(root):
         root = download(root)
-    rpath = Path(root).expanduser()
 
+    rpath = Path(root).expanduser()
     labels_file = os.path.join(rpath, "labels.txt")
+
     if os.path.exists(labels_file):
         with open(labels_file) as f:
             class_names = [line.rstrip().lower() for line in f]
         print(f"using class_names in labels.txt: {class_names}")
     else:
-        ## read the class names and save results
+        # read the class names and save results
         logger.warning(
             "labels.txt does not exist, using default VOC names. "
             "Creating labels.txt by scanning the directory: {}".format(os.path.join(root, "Annotations"))
@@ -1223,28 +1268,46 @@ def get_voc_format_classes(root: str):
             voc_annotation_path=os.path.join(root, "Annotations"), voc_class_names_output_path=labels_file
         )
 
-    return class_names
+    # There are no category IDs in VOC format
+    # Create category IDs dictionary starting from 1
+    # 0 is typically reserved for background in many frameworks
+    category_ids = [idx + 1 for idx, name in enumerate(class_names)]
+
+    return class_names, category_ids
 
 
 def get_detection_classes(sample_data_path):
     """
-    The all class names for given data.
+    Get class names and category IDs from detection dataset in various formats.
 
     Parameters
     ----------
-    sample_data_path
-        If it is a file, it is the path to COCO format json annotation file. Could be any split, e.g. train/val/test/....
-        If it is a directory, it is path to the root directory of VOC data.
+    sample_data_path : Union[str, pd.DataFrame]
+        The input can be one of:
+        - str (directory): Path to root directory of VOC format data
+        - str (file): Path to COCO format JSON annotation file
+        - pd.DataFrame: DataFrame containing detection data with 'rois' column
 
     Returns
     -------
-        All the class names.
+    tuple
+        A tuple containing (class_names, category_ids) where:
+        - class_names: list of class name strings
+        - category_ids: dict mapping class names to their numeric IDs
+
+        For VOC: IDs start from 1
+        For COCO: Original category IDs from annotation file
+        For DataFrame: Sequential IDs starting from 1
     """
+    # Handle string paths for VOC and COCO formats
     if isinstance(sample_data_path, str):
         if os.path.isdir(sample_data_path):
+            # Directory path indicates VOC format
             return get_voc_format_classes(sample_data_path)
         else:
+            # File path indicates COCO format JSON
             return get_coco_format_classes(sample_data_path)
+    # Handle DataFrame format
     elif isinstance(sample_data_path, pd.DataFrame):
         return get_df_unique_classes(sample_data_path)
 
@@ -1492,7 +1555,7 @@ def get_color(idx):
     return color
 
 
-def save_result_df(
+def convert_result_df(
     pred: Iterable, data: Union[pd.DataFrame, Dict], detection_classes: List[str], result_path: Optional[str] = None
 ):
     """
@@ -1539,11 +1602,11 @@ def save_result_df(
     return result_df
 
 
-def save_result_coco_format(detection_data_path, pred, result_path):
-    coco_dataset = COCODataset(detection_data_path)
+def save_result_coco_format(data_path, pred, category_ids, result_path, coco_root: Optional[str] = None):
+    coco_dataset = COCODataset(data_path, category_ids=category_ids)
     result_name, _ = os.path.splitext(result_path)
     result_path = result_name + ".json"
-    coco_dataset.save_result(pred, from_coco_or_voc(detection_data_path, "test"), save_path=result_path)
+    coco_dataset.save_result(pred, from_coco_or_voc(data_path, "test", coco_root=coco_root), save_path=result_path)
     logger.info(25, f"Saved detection result to {result_path}")
 
 
@@ -1554,9 +1617,30 @@ def save_result_voc_format(pred, result_path):
     logger.info(25, f"Saved detection result to {result_path}")
 
 
-def convert_pred_to_xywh(pred: Optional[List]):
+def convert_pred_to_xywh(pred: Optional[List]) -> Optional[List]:
+    """
+    Convert prediction bounding boxes from XYXY to XYWH format.
+
+    Args:
+        pred: List of predictions, where each prediction contains 'bboxes' that can be
+             either a torch.Tensor or numpy.ndarray
+
+    Returns:
+        Modified list of predictions with bboxes in XYWH format
+    """
     if not pred:
         return pred
+
     for i, pred_per_image in enumerate(pred):
-        pred[i]["bboxes"] = bbox_xyxy_to_xywh(pred_per_image["bboxes"].detach().numpy())
+        bboxes = pred_per_image["bboxes"]
+
+        # Handle numpy array case
+        if isinstance(bboxes, np.ndarray):
+            pred[i]["bboxes"] = bbox_xyxy_to_xywh(bboxes)
+        # Handle torch.Tensor case
+        elif torch.is_tensor(bboxes):
+            pred[i]["bboxes"] = bbox_xyxy_to_xywh(bboxes.detach().numpy())
+        else:
+            raise TypeError(f"Unsupported bbox type: {type(bboxes)}. Expected numpy.ndarray or torch.Tensor")
+
     return pred

autogluon/multimodal/version.py

@@ -1,3 +1,3 @@
 """This is the autogluon version file."""
-__version__ = '1.1.2b20241106'
+__version__ = '1.1.2b20241108'
 __lite__ = False

{autogluon.multimodal-1.1.2b20241106.dist-info → autogluon.multimodal-1.1.2b20241108.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.multimodal
-Version: 1.1.2b20241106
+Version: 1.1.2b20241108
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -41,10 +41,10 @@ Requires-Dist: scikit-learn<1.5.3,>=1.4.0
 Requires-Dist: Pillow<12,>=10.0.1
 Requires-Dist: tqdm<5,>=4.38
 Requires-Dist: boto3<2,>=1.10
-Requires-Dist: torch<2.5,>=2.2
-Requires-Dist: lightning<2.4,>=2.2
-Requires-Dist: transformers[sentencepiece]<4.41.0,>=4.38.0
-Requires-Dist: accelerate<0.34.0,>=0.21.0
+Requires-Dist: torch<2.6,>=2.2
+Requires-Dist: lightning<2.6,>=2.2
+Requires-Dist: transformers[sentencepiece]<5,>=4.38.0
+Requires-Dist: accelerate<1.0,>=0.32.0
 Requires-Dist: requests<3,>=2.30
 Requires-Dist: jsonschema<4.22,>=4.18
 Requires-Dist: seqeval<1.3.0,>=1.2.2
@@ -56,9 +56,9 @@ Requires-Dist: text-unidecode<1.4,>=1.3
 Requires-Dist: torchmetrics<1.3.0,>=1.2.0
 Requires-Dist: nptyping<2.5.0,>=1.4.4
 Requires-Dist: omegaconf<2.3.0,>=2.1.1
-Requires-Dist: autogluon.core[raytune]==1.1.2b20241106
-Requires-Dist: autogluon.features==1.1.2b20241106
-Requires-Dist: autogluon.common==1.1.2b20241106
+Requires-Dist: autogluon.core[raytune]==1.1.2b20241108
+Requires-Dist: autogluon.features==1.1.2b20241108
+Requires-Dist: autogluon.common==1.1.2b20241108
 Requires-Dist: pytorch-metric-learning<2.4,>=1.3.0
 Requires-Dist: nlpaug<1.2.0,>=1.1.10
 Requires-Dist: nltk<3.9,>=3.4.5

{autogluon.multimodal-1.1.2b20241106.dist-info → autogluon.multimodal-1.1.2b20241108.dist-info}/RECORD

@@ -1,11 +1,11 @@
-autogluon.multimodal-1.1.2b20241106-py3.8-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
+autogluon.multimodal-1.1.2b20241108-py3.8-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
 autogluon/multimodal/__init__.py,sha256=EuWb-QmtFBKePJw4_4Kpp9dKrabv121haYw_Oiu2jfI,238
 autogluon/multimodal/constants.py,sha256=lFA03ZQeZlp8mwuXLuMOgeByljV5wPYBjN_hi6Xc8zg,8498
-autogluon/multimodal/predictor.py,sha256=VTJGcKH4Kktdm4Qq2x9oRThpfp6w_kFSjJOmQPsmB5g,40654
+autogluon/multimodal/predictor.py,sha256=beV2gOcTnviYtU8UWTWdqWYTbuk5sC6Sba-pAEaFQyg,40936
 autogluon/multimodal/presets.py,sha256=VR_arn7X4eiQcGcvJVmwxDopPJGvYP1W1cBZ2AOcdJM,25882
 autogluon/multimodal/problem_types.py,sha256=JPSoV3Fg-NGQansRlyT2bPZG3iIkgWo7eB8oDoAfg90,9201
 autogluon/multimodal/registry.py,sha256=vqvONWweZP44aBo1jCvlqLdL0Agn2kyTK8uXUwagZhs,3670
-autogluon/multimodal/version.py,sha256=ZUeAMVbkYXpcl_SBazx8fDZGfJkzecMES3KYm_c1OxU,90
+autogluon/multimodal/version.py,sha256=LF2xpVDjwJNbbPJVJI9mLUJAKTcMv21XRyuIrXDNfTI,90
 autogluon/multimodal/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autogluon/multimodal/cli/prepare_detection_dataset.py,sha256=9NCYmCUMPRWbxxbN_C7YQjMYlrAm8gbwJ3Qd-79JWH4,5218
 autogluon/multimodal/cli/voc2coco.py,sha256=LXVu9RIfOZs_1URrzO-_3Nrvb9uGEgPxCY4-B6m1coc,9605
@@ -14,14 +14,14 @@ autogluon/multimodal/configs/data/default.yaml,sha256=gsadTDJ3ZbppQ5rpA0sPhqd9Mw
 autogluon/multimodal/configs/distiller/default.yaml,sha256=DiCZYYJDEk5k03ZI-ewj9hWiKpMQbB8oqjWYavMK1wU,513
 autogluon/multimodal/configs/environment/default.yaml,sha256=Di1EohFWUoGi0bLuJfOr3bDD4ceD_ZrfDc_GJO55mpk,1235
 autogluon/multimodal/configs/matcher/default.yaml,sha256=K0ehM0uIFfKq1CeeaFcv14RBjo3khMgWKS2ymWI-V9I,218
-autogluon/multimodal/configs/model/default.yaml,sha256=7cLxvza8AEfWEyLfrXe89kn1pvaY5JsvgBBB38vi_xI,8731
+autogluon/multimodal/configs/model/default.yaml,sha256=NgCPV01hfcG21hvmtLaiTbahTYkRi3ZmHONzsRdwjOE,8751
 autogluon/multimodal/configs/optimization/default.yaml,sha256=w4xlVS1gUK8tDXyhfDtIKabTqr_gBWSduq1y8T9CBI8,2182
 autogluon/multimodal/configs/pretrain/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autogluon/multimodal/configs/pretrain/detection/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autogluon/multimodal/configs/pretrain/detection/coco_detection.py,sha256=UlSwkWAkST_96RTzPZMuPuqIfv72U03-JdqwPd-NjiQ,3171
 autogluon/multimodal/configs/pretrain/detection/default_runtime.py,sha256=9hJmjxb6fIo-kbbejQlJy4ayopRFUyA_w95plhAUFDw,793
 autogluon/multimodal/configs/pretrain/detection/schedule_1x.py,sha256=VhZ8HT-ryeGW-GzxiVsDEIYf9Bw6ImOdPucFVJaN0Os,298
-autogluon/multimodal/configs/pretrain/detection/dino/dino-4scale_r50_8xb2-12e_coco.py,sha256=Zm41N-VNY13cj6U485wtrDqB7k49Sdj4MdNLYjTE62E,5910
+autogluon/multimodal/configs/pretrain/detection/dino/dino-4scale_r50_8xb2-12e_coco.py,sha256=r51gu03JBHrTjDXkT2CdNKTzEeWx2qxk8t-TEHgKwSQ,5919
 autogluon/multimodal/configs/pretrain/detection/dino/dino-5scale_swin-l_8xb2-12e_coco.py,sha256=VYvjELfzX2jYmsQLFvMYWeixx2LmB_LwbPmzgLAPUwo,1137
 autogluon/multimodal/configs/pretrain/detection/dino/dino-5scale_swin-l_8xb2-36e_coco.py,sha256=OL4Za_hd5IhQU8iHEAVrsFu-MnSmOPu1_WRa39i3QYA,266
 autogluon/multimodal/configs/pretrain/detection/dino/dino_swinl_tta.py,sha256=vi5rhbaT3mgycIm1W8jQ6l-KuFoL24OiiF-rdj2CTNg,68
@@ -71,7 +71,7 @@ autogluon/multimodal/learners/base.py,sha256=IUHRBzwrKqAwo9nDsqzg0rBQaFiVxfyeMnd
 autogluon/multimodal/learners/few_shot_svm.py,sha256=TXauhXr_2hWqaEwO8XhFxWRRPXDYxLpmmKYaCrxFWPM,23934
 autogluon/multimodal/learners/matching.py,sha256=gueWrqy0g9gVbyBvQOAj03JgBwbJsBXeLLtKsiTzrnU,89891
 autogluon/multimodal/learners/ner.py,sha256=0VZl_Z1O98A5mOSw8Ee8F9foaCT684DT0j1ALx-8RU4,19035
-autogluon/multimodal/learners/object_detection.py,sha256=JRxXaI33Um0BWGsRRKFfLy7tAfqJj2SsT_DqIL1igKo,29015
+autogluon/multimodal/learners/object_detection.py,sha256=NmaQWaakpnIA1owZNvGeO1B3Sk3p3ngWnNJM21EeeVM,30846
 autogluon/multimodal/learners/semantic_segmentation.py,sha256=cy2ALYTtqeSnPo75htgr5STZ_utgkzeGxp5j4J1mScc,20183
 autogluon/multimodal/models/__init__.py,sha256=wynO5U5zY_vElZPGL10Oqb7OVkj2iovqzml22MRL3iE,842
 autogluon/multimodal/models/adaptation_layers.py,sha256=NuzwU_ghk8D2axmDuD8UEZ_HamoMSCcKMV9DB1AYWAg,38425
@@ -109,7 +109,7 @@ autogluon/multimodal/optimization/losses.py,sha256=n1nXpXgGmYfLv0b-51yDFp99szy3j
 autogluon/multimodal/optimization/lr_scheduler.py,sha256=i3GG7T8ZyPXyS7feUVe7W3o6eSLIh_Ei7XujJL50uxw,5829
 autogluon/multimodal/optimization/semantic_seg_metrics.py,sha256=tIbSk3iyBRRx7HnZdqIxltRBtDiBt-GX_zBxkMOFxQg,32894
 autogluon/multimodal/optimization/utils.py,sha256=X6UknHcN2XID9WO2N2Of3-7MbWUfZO7ydNOktwwtbiU,34415
-autogluon/multimodal/utils/__init__.py,sha256=SaUQE-TwodG9NVOchXGp-Fx1CKKZe9iUKhFD_cTiI0c,2883
+autogluon/multimodal/utils/__init__.py,sha256=GQiubE17Z5FVG7o5kZgwekmTHkYd6aL1MliZF21kbFI,2886
 autogluon/multimodal/utils/cache.py,sha256=XiLB5xNUYklDc8R9M-2RUD0Y6NEqrZIZx6O1PpRIXxI,7766
 autogluon/multimodal/utils/checkpoint.py,sha256=Khx4KXqF0S9Aw193kyUWNvJX2XkFv6y4IGYe9-txLJE,7699
 autogluon/multimodal/utils/cloud_io.py,sha256=FhIJ_oEerJ8QOZz82SzP3tg-Z1mo_vHqlmcShHe077s,3031
@@ -120,7 +120,7 @@ autogluon/multimodal/utils/distillation.py,sha256=VYSY_excOESa84Q0w6IbV4wL_ER27W
 autogluon/multimodal/utils/download.py,sha256=eOCw4b_EHBvHB9EcGesyzTm1f2AUrzz8KcZQ6j_D7-Y,10364
 autogluon/multimodal/utils/environment.py,sha256=J1YYBcIL-YzAnoN5GC1DMF_Rpt0DxSpJp3NZxJZ_q6g,12814
 autogluon/multimodal/utils/export.py,sha256=h7PizWsMaxMnlY7ssRNJxbExgt7B4XFkKY8hf7M-j4Y,11964
-autogluon/multimodal/utils/hpo.py,sha256=auaeeksOsDFCFWVnb0uFapNt4pZjGv8RC3fEO-r5MuU,8717
+autogluon/multimodal/utils/hpo.py,sha256=KIpO7DmvowT-xua78yK5akKOFrJFKDLQajwaFkX0WWU,8813
 autogluon/multimodal/utils/inference.py,sha256=VQAda55sf6rbuuxUS3MGJXyCAgb_xugLv1glCV2NlZk,12349
 autogluon/multimodal/utils/label_studio.py,sha256=7lFl75zztIy6VCuCbyZkN-BLbtr0j1S4F42zJteGVYY,13437
 autogluon/multimodal/utils/load.py,sha256=rzfADn6obbZL20QZc034IPhIiza7SA8f5MPpd9hPsAE,5106
@@ -131,15 +131,15 @@ autogluon/multimodal/utils/misc.py,sha256=WaDWN-6xCCL4tCkxMr4VMb5oiNmmBLrWo5FC3b
 autogluon/multimodal/utils/mmcv.py,sha256=Jjg5PiPqiRNJk6yWkQQlNiqT7qhStN94QjqQsZO3uVw,922
 autogluon/multimodal/utils/model.py,sha256=Z_9bev8nRk92cUZjPggVAM3R3CHFlecU-YzjkMGPUsE,21963
 autogluon/multimodal/utils/nlpaug.py,sha256=rWztiOZDbtEGBdyjkXZ0DoSEpXGKX9akpDpFnz4xIMw,2557
-autogluon/multimodal/utils/object_detection.py,sha256=EjLPzmq8Ge_HAtibiY5FNOChL_qAIls4PJF3PFd8kYA,50047
+autogluon/multimodal/utils/object_detection.py,sha256=fHZxon5LoYRmNu_7jm_pDjesVxTa72nzZwgwP-5Fft8,53535
 autogluon/multimodal/utils/onnx.py,sha256=rblWnphKTsfbosbieJu8PsH6SMDw4on9BS8bR1plL2U,5607
 autogluon/multimodal/utils/save.py,sha256=zYIO3mYMGBvHfZcmCUaLpsQa14nVq1LPv2F76uaz89w,3951
 autogluon/multimodal/utils/visualizer.py,sha256=qAc4_36r3MBDPq1R1jBeb_gcfzIrsylL1S31sRf3wuw,22562
-autogluon.multimodal-1.1.2b20241106.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
-autogluon.multimodal-1.1.2b20241106.dist-info/METADATA,sha256=OCuGdZ0aV9Ifhx2E1BqKANqtzQDp1Qkl5Q2bc1d8BMk,12888
-autogluon.multimodal-1.1.2b20241106.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
-autogluon.multimodal-1.1.2b20241106.dist-info/WHEEL,sha256=eOLhNAGa2EW3wWl_TU484h7q1UNgy0JXjjoqKoxAAQc,92
-autogluon.multimodal-1.1.2b20241106.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.multimodal-1.1.2b20241106.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.multimodal-1.1.2b20241106.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
-autogluon.multimodal-1.1.2b20241106.dist-info/RECORD,,
+autogluon.multimodal-1.1.2b20241108.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
+autogluon.multimodal-1.1.2b20241108.dist-info/METADATA,sha256=6I1NONkJaNxI-zUF9uUZZChGOW6UwDAaFvuhej3tF0E,12880
+autogluon.multimodal-1.1.2b20241108.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
+autogluon.multimodal-1.1.2b20241108.dist-info/WHEEL,sha256=eOLhNAGa2EW3wWl_TU484h7q1UNgy0JXjjoqKoxAAQc,92
+autogluon.multimodal-1.1.2b20241108.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.multimodal-1.1.2b20241108.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.multimodal-1.1.2b20241108.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+autogluon.multimodal-1.1.2b20241108.dist-info/RECORD,,