boris-behav-obs 9.7.12__py3-none-any.whl → 9.7.15__py3-none-any.whl

boris/analysis_plugins/_export_to_feral.py

@@ -1,225 +0,0 @@
-"""
-BORIS plugin
-
-Export to FERAL (getferal.ai)
-"""
-
-import pandas as pd
-import json
-from pathlib import Path
-
-from PySide6.QtWidgets import QFileDialog
-
-# dependencies for CategoryDialog
-from PySide6.QtWidgets import QListWidget, QListWidgetItem, QLabel, QPushButton, QVBoxLayout, QHBoxLayout, QDialog
-from PySide6.QtCore import Qt
-
-
-__version__ = "0.1.1"
-__version_date__ = "2025-11-28"
-__plugin_name__ = "Export observations to FERAL"
-__author__ = "Olivier Friard - University of Torino - Italy"
-
-
-class CategoryDialog(QDialog):
-    def __init__(self, items, parent=None):
-        super().__init__(parent)
-
-        self.setWindowTitle("Organize the videos in categories")
-
-        self.setModal(True)
-
-        # Main layout
-        main_layout = QVBoxLayout(self)
-        lists_layout = QHBoxLayout()
-
-        # All videos
-        self.list_unclassified = self._create_list_widget()
-        self.label_unclassified = QLabel("All videos")
-        col0_layout = QVBoxLayout()
-        col0_layout.addWidget(self.label_unclassified)
-        col0_layout.addWidget(self.list_unclassified)
-
-        self.list_cat1 = self._create_list_widget()
-        self.label_cat1 = QLabel("train")
-        col1_layout = QVBoxLayout()
-        col1_layout.addWidget(self.label_cat1)
-        col1_layout.addWidget(self.list_cat1)
-
-        self.list_cat2 = self._create_list_widget()
-        self.label_cat2 = QLabel("val")
-        col2_layout = QVBoxLayout()
-        col2_layout.addWidget(self.label_cat2)
-        col2_layout.addWidget(self.list_cat2)
-
-        self.list_cat3 = self._create_list_widget()
-        self.label_cat3 = QLabel("test")
-        col3_layout = QVBoxLayout()
-        col3_layout.addWidget(self.label_cat3)
-        col3_layout.addWidget(self.list_cat3)
-
-        self.list_cat4 = self._create_list_widget()
-        self.label_cat4 = QLabel("inference")
-        col4_layout = QVBoxLayout()
-        col4_layout.addWidget(self.label_cat4)
-        col4_layout.addWidget(self.list_cat4)
-
-        # Add all columns to the horizontal layout
-        lists_layout.addLayout(col0_layout)
-        lists_layout.addLayout(col1_layout)
-        lists_layout.addLayout(col2_layout)
-        lists_layout.addLayout(col3_layout)
-        lists_layout.addLayout(col4_layout)
-
-        main_layout.addLayout(lists_layout)
-
-        buttons_layout = QHBoxLayout()
-        self.btn_ok = QPushButton("OK")
-        self.btn_cancel = QPushButton("Cancel")
-
-        self.btn_ok.clicked.connect(self.accept)
-        self.btn_cancel.clicked.connect(self.reject)
-
-        buttons_layout.addStretch()
-        buttons_layout.addWidget(self.btn_ok)
-        buttons_layout.addWidget(self.btn_cancel)
-
-        main_layout.addLayout(buttons_layout)
-
-        # Populate "Unclassified" with input items
-        for text in items:
-            QListWidgetItem(text, self.list_unclassified)
-
-    def _create_list_widget(self):
-        """
-        Create a QListWidget ready for drag & drop.
-        """
-        lw = QListWidget()
-        lw.setSelectionMode(QListWidget.ExtendedSelection)
-        lw.setDragEnabled(True)
-        lw.setAcceptDrops(True)
-        lw.setDropIndicatorShown(True)
-        lw.setDragDropMode(QListWidget.DragDrop)
-        lw.setDefaultDropAction(Qt.MoveAction)
-        return lw
-
-    def get_categories(self):
-        """
-        Return the content of all categories as a dictionary of lists.
-        """
-
-        def collect(widget):
-            return [widget.item(i).text().rstrip("*") for i in range(widget.count())]
-
-        return {
-            "unclassified": collect(self.list_unclassified),
-            "train": collect(self.list_cat1),
-            "val": collect(self.list_cat2),
-            "test": collect(self.list_cat3),
-            "inference": collect(self.list_cat4),
-        }
-
-
-def run(df: pd.DataFrame, project: dict):
-    """
-    Export observations to FERAL
-    See https://www.getferal.ai/ > Label Preparation
-    """
-
-    out: dict = {
-        "is_multilabel": False,
-        "splits": {
-            "train": [],
-            "val": [],
-            "test": [],
-            "inference": [],
-        },
-    }
-
-    log: list = []
-
-    # class names
-    class_names = {x: project["behaviors_conf"][x]["code"] for x in project["behaviors_conf"]}
-    out["class_names"] = class_names
-    reversed_class_names = {project["behaviors_conf"][x]["code"]: int(x) for x in project["behaviors_conf"]}
-    log.append(f"{class_names=}")
-
-    observations: list = sorted([x for x in project["observations"]])
-    log.append(f"Selected observation: {observations}")
-
-    labels: dict = {}
-    video_list: list = []
-    for observation_id in observations:
-        log.append("---")
-        log.append(observation_id)
-
-        # check number of media file in player #1
-        if len(project["observations"][observation_id]["file"]["1"]) != 1:
-            log.append(f"The observation {observation_id} contains more than one video")
-            continue
-
-        # check number of coded subjects
-        if len(set([x[1] for x in project["observations"][observation_id]["events"]])) > 1:
-            log.append(f"The observation {observation_id} contains more than one subject")
-            continue
-
-        media_file_path: str = project["observations"][observation_id]["file"]["1"][0]
-        media_file_name = str(Path(media_file_path).name)
-
-        # skip if no events
-        if not project["observations"][observation_id]["events"]:
-            video_list.append(media_file_name)
-            log.append(f"No events for observation {observation_id}")
-            continue
-        else:
-            video_list.append(media_file_name + "*")
-
-        # extract FPS
-        FPS = project["observations"][observation_id]["media_info"]["fps"][media_file_path]
-        log.append(f"{media_file_name} {FPS=}")
-        # extract media duration
-        duration = project["observations"][observation_id]["media_info"]["length"][media_file_path]
-        log.append(f"{media_file_name} {duration=}")
-
-        number_of_frames = int(duration / (1 / FPS))
-        log.append(f"{number_of_frames=}")
-
-        labels[media_file_name] = [0] * number_of_frames
-
-        for idx in range(number_of_frames):
-            t = idx * (1 / FPS)
-            behaviors = (
-                df[(df["Observation id"] == observation_id) & (df["Start (s)"] <= t) & (df["Stop (s)"] >= t)]["Behavior"].unique().tolist()
-            )
-            if len(behaviors) > 1:
-                log.append(f"The observation {observation_id} contains more than one behavior for frame {idx}")
-                del labels[media_file_name]
-                break
-            if behaviors:
-                behaviors_idx = reversed_class_names[behaviors[0]]
-                labels[media_file_name][idx] = behaviors_idx
-
-    out["labels"] = labels
-
-    # splits
-    dlg = CategoryDialog(video_list)
-
-    if dlg.exec():  # Dialog accepted
-        result = dlg.get_categories()
-        del result["unclassified"]
-        out["splits"] = result
-
-        filename, _ = QFileDialog.getSaveFileName(
-            None,
-            "Choose a file to save",
-            "",  # start directory
-            "JSON files (*.json);;All files (*.*)",
-        )
-        if filename:
-            with open(filename, "w") as f_out:
-                f_out.write(json.dumps(out, separators=(",", ": "), indent=1))
-
-    else:
-        log.append("splits section missing")
-
-    return "\n".join(log)