britekit 0.0.11__py3-none-any.whl → 0.0.12__py3-none-any.whl

britekit/__about__.py

@@ -1,4 +1,4 @@
 # SPDX-FileCopyrightText: 2025-present Jan Huus <jhuus1@gmail.com>
 #
 # SPDX-License-Identifier: MIT
-__version__ = "0.0.11"
+__version__ = "0.0.12"

britekit/commands/_xeno.py

@@ -4,6 +4,7 @@ import json
 import logging
 import os
 from typing import Optional
+from urllib.parse import quote
 
 import click
 
@@ -101,6 +102,8 @@ def xeno(
     else:
         name = f'en:"={name.lower()}"'
 
+    name = quote(name)
+
     # get list of recordings
     recordings: list = []
     page = 0
@@ -137,6 +140,9 @@ def xeno(
         if not os.path.exists(outfile):
             logging.info(f"Downloading {outfile}")
             url = recording["file"]
+            if not url:
+                url = f"https:{recording['url']}/download"
+
             response = requests.get(url)
             with open(outfile, "wb") as mp3:
                 mp3.write(response.content)

britekit/core/audio.py

@@ -264,7 +264,6 @@ class Audio:
         import numpy as np
         import torch
 
-        """Create a filterbank for log spectrograms."""
         f_min = self.cfg.audio.min_freq
         f_max = self.cfg.audio.max_freq
         n_bins = self.cfg.audio.spec_height
@@ -282,13 +281,19 @@ class Audio:
 
         filters = []
         for log_cf in log2_centers:
+            # Gaussian filter in log2(f)
             weight = np.exp(-0.5 * ((fft_log2 - log_cf) / sigma_log2) ** 2)
-            weight_sum = np.sum(weight)
-            if weight_sum > 0:
-                filters.append(weight / weight_sum)
-            else:
-                # Fallback: use uniform weights if sum is zero
-                filters.append(weight)
+
+            # Normalize per filter (contrast-preserving)
+            weight /= np.sum(weight) + 1e-12
+
+            # Boost power in higher frequencies, so more like mel scale;
+            # increase log_freq_gain to increase the boost
+            cf_hz = 2**log_cf
+            gain = (cf_hz / f_min) ** self.cfg.audio.log_freq_gain
+            weight *= gain
+
+            filters.append(weight)
 
         filters = np.array(filters)
         return torch.tensor(filters, dtype=torch.float32).to(self.device)
@@ -323,28 +328,41 @@ class Audio:
         tensor = torch.from_numpy(signal).to(self.device)
 
         if freq_scale == "log":
-            spec = self.linear_transform(tensor)
-            spec = torch.matmul(self.log2_filterbank, spec.squeeze(0)).unsqueeze(0)
+            spec = self.linear_transform(tensor)  # [1, n_freqs, n_frames]
+            spec = torch.matmul(
+                self.log2_filterbank, spec.squeeze(0)
+            )  # [n_mels, n_frames]
+            spec = spec.unsqueeze(0).unsqueeze(1)  # [1, 1, n_mels, n_frames]
+
         elif freq_scale == "mel":
-            spec = self.mel_transform(tensor)
+            spec = self.mel_transform(tensor).unsqueeze(1)  # [1, 1, n_mels, T]
+
         elif freq_scale == "linear":
             spec = self.linear_transform(tensor)
-
             freqs = torch.fft.rfftfreq(
                 2 * self.win_length, d=1 / self.cfg.audio.sampling_rate
-            )  # [freq_bins]
+            )
             mask = (freqs >= self.cfg.audio.min_freq) & (
                 freqs <= self.cfg.audio.max_freq
             )
-            spec = spec[:, mask, :]  # shape: [channel, selected_freq_bins, time_frames]
-            spec = spec.unsqueeze(1)
-            spec = F.interpolate(
-                spec,
-                size=(self.cfg.audio.spec_height, self.cfg.audio.spec_width),
-                mode="bilinear",
-                align_corners=False,
-            )
-            spec = spec.squeeze(1)
+            spec = spec[:, mask, :].unsqueeze(1)  # [1, 1, F_sel, T]
+
+        # downsample frequency to spec_height (energy-preserving)
+        spec = F.interpolate(
+            spec,
+            size=(self.cfg.audio.spec_height, spec.shape[-1]),
+            mode="area",
+        )
+
+        # pad or crop to spec_width
+        T = spec.shape[-1]
+        if T < self.cfg.audio.spec_width:
+            pad_width = self.cfg.audio.spec_width - T
+            spec = F.pad(spec, (0, pad_width))  # pad on the right (time axis)
+        else:
+            spec = spec[..., : self.cfg.audio.spec_width]
+
+        spec = spec.squeeze(1)
 
         if decibels:
             spec = ta.transforms.AmplitudeToDB(stype="power", top_db=top_db)(spec)

britekit/core/base_config.py

@@ -23,6 +23,7 @@ class Audio:
     decibels: bool = False  # Use decibel amplitude scale?
     top_db: float = 80  # Parameter to decibel conversion
     db_power: float = 1.0  # Raise to this exponent after convert to decibels
+    log_freq_gain: float = 0.6  # Boost loudness of higher frequencies with log scale
 
 
 @dataclass

britekit/core/trainer.py

@@ -47,7 +47,9 @@ class Trainer:
         val_rocs = []
         for k in range(self.cfg.train.num_folds):
             logger = TensorBoardLogger(
-                save_dir="logs", name=f"fold-{k}", default_hp_metric=False
+                save_dir="logs",
+                name=None if self.cfg.train.num_folds == 1 else f"fold-{k}",
+                default_hp_metric=False
             )
             version = (
                 logger.version

britekit/install/yaml/base_config.yaml

@@ -5,6 +5,7 @@ audio: !!python/object:britekit.core.base_config.Audio
   db_power: 1.0
   decibels: false
   freq_scale: mel
+  log_freq_gain: 0.6
   max_freq: 8000
   min_freq: 100
   power: 1.0

britekit/models/base_model.py

@@ -177,7 +177,7 @@ class BaseModel(pl.LightningModule):
         else:
             preds = torch.softmax(seg_logits, dim=1)
 
-        self.log("val_loss", loss, on_step=False, on_epoch=True, prog_bar=True)
+        self.log("val_loss", loss, on_step=False, on_epoch=True, prog_bar=False)
 
         roc_auc = metrics.roc_auc_score(y.cpu(), preds.cpu(), average="micro")
         self.log("val_roc", roc_auc, on_step=False, on_epoch=True, prog_bar=True)

{britekit-0.0.11.dist-info → britekit-0.0.12.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: britekit
-Version: 0.0.11
+Version: 0.0.12
 Summary: Core functions for bioacoustic recognizers.
 Project-URL: Documentation, https://github.com/jhuus/BriteKit#readme
 Project-URL: Issues, https://github.com/jhuus/BriteKit/issues

{britekit-0.0.11.dist-info → britekit-0.0.12.dist-info}/RECORD

@@ -1,8 +1,8 @@
 britekit/cli.py,sha256=nnrCMfw3-1GJ4rKFpqTLu8JcBGxTocMn7nwzU4OSaew,3080
 britekit/core/analyzer.py,sha256=0OYVxuwYx_R36MZwIlVgPMWQ1udZ_tdgYODJyEVDJm0,5604
-britekit/core/audio.py,sha256=Ju1SBO9c8GUM0RZ-NuolZoPHF6hATLvayVEvKVOgxOw,15850
+britekit/core/audio.py,sha256=2WINXunumjcbFlTd0Yngg8MmBy9O3dQWGosVCJmkryI,16397
 britekit/core/augmentation.py,sha256=5_wyB-6gt7uM68Zl-rO_fPu1D6tlsd2m5oWhA6l0W9Q,5721
-britekit/core/base_config.py,sha256=wbCJI9cEH9mktUTSfLSCnU5AhJT6xhxhwZS4QWRYTpM,8744
+britekit/core/base_config.py,sha256=KNcmPhvbl6OGDZ5OvHShE9JCnuyhSvIQq5pi3AU1gVk,8830
 britekit/core/config_loader.py,sha256=mDmLwKYXsWb9FRk12q9Rj2rVdvbCVCI3QTdsZSnUBCY,1339
 britekit/core/data_module.py,sha256=0DkOZTTGymZ7JjukucDuAq-nXS_KPhqV7KFPhmMoQDw,9105
 britekit/core/dataset.py,sha256=R_NDaFljl4IMKxGWWWm1PMkYCcfrEaiJU1NrVswt6eM,5801
@@ -11,10 +11,10 @@ britekit/core/pickler.py,sha256=72hiqOiIbBeXvHGwmHAOODr9wGdyBwj4NUYecxIhtCE,5775
 britekit/core/plot.py,sha256=H-SB2ooe2LT0j1KoXs2PgT3t05oAB4CWkYX_HvGMO_c,5329
 britekit/core/predictor.py,sha256=9Heue3ku9rw4OfE0SWuWLKBLW-aMw8PXqWlBB4GUCbw,22911
 britekit/core/reextractor.py,sha256=UqWLapReNhEne0dykBCN_UIpLOHkZxJMw4RAlG70nzI,8393
-britekit/core/trainer.py,sha256=vFIVyFCYhnV0zdRjNU93wzd7_HLtiaXW8mAiK7BoJPA,6437
+britekit/core/trainer.py,sha256=XCa4yOkKeX2CBPPyH3gsJBalhrsebRboH8EC4rLdceA,6512
 britekit/core/tuner.py,sha256=LoQi7jwIwqv2DY7MgupG32PdD6QiuneQrgRgxq8ANlk,16422
 britekit/core/util.py,sha256=JKT-yqFPA9FBMT1-YsGvNE8E21242MYAsXvRGSgTNYg,19163
-britekit/models/base_model.py,sha256=VpY2aMb7P3QWGzhIdJRGxdjT6z-o7XE5sJxnupeZtxQ,16880
+britekit/models/base_model.py,sha256=jVYSIrXAmeAer6MTFg177YoDBn00_HzluDdEAZH1qpQ,16881
 britekit/models/dla.py,sha256=ALMY997AbERN7-sHqQuE5e43llRjpUDPZSFGL-Flv4M,3137
 britekit/models/effnet.py,sha256=e7WdZMsLPXe8jcWChk6n97c8DMV0YyGV6lDP_Jv6Wz4,3129
 britekit/models/gernet.py,sha256=7MEUZaDTfr-6oa8eE8dyDQb2LgahGBOEp1pTZSu1KOE,7022
@@ -32,7 +32,7 @@ britekit/testing/per_segment_tester.py,sha256=5zFpe8wVEEtVjthKKFRb0dwE9T0v4JIll9
 britekit/training_db/extractor.py,sha256=yblYTco9b-bZhBTOkGkNANOabsNo8IfQGJxPvh_eJyo,8406
 britekit/training_db/training_data_provider.py,sha256=V5aBjsCvrWViZ0Jv05hgcKRizcAXmqoj4q3hAHedoD8,5651
 britekit/training_db/training_db.py,sha256=xZqN1sMC2yFMEDm9rOrCigN3CUfUzTaTUkX3IZ_zHRc,64922
-britekit/__about__.py,sha256=CGewA3qxQM5QGFXz6GxYj1rsR4YlBI3pPly5AWErQFI,123
+britekit/__about__.py,sha256=BYHVclZkphegQAtQMaCBTllSQbN2DGte-0mtbt9t5SM,123
 britekit/__init__.py,sha256=RpruzdjbvTcFNf21zJYY8HrAhJei91FtNNLjIBmw-kw,1857
 britekit/install/data/classes.csv,sha256=OdTZ8oQdx7N-HKyhftxZStGZYsjhCy4UbanwtQJ2wBM,54
 britekit/install/data/ignore.txt,sha256=RbKvEHtUCbgRYolwR1IucClwyD3q7l2s6QuRjph-Us4,68
@@ -52,7 +52,7 @@ britekit/install/data/audioset/curated/siren.csv,sha256=_EbIBGHJPsR7VoA0gEYG7DCc
 britekit/install/data/audioset/curated/speech.csv,sha256=eBGVyfFTPHr34XrEIQIy3CWQfXMjQP5JqjryRI3IF8c,2439
 britekit/install/data/audioset/curated/truck.csv,sha256=i0cTXlqGQwv2ULnrZTJ68p3EV8UWccU0saw7fqQq_8E,1667
 britekit/install/data/audioset/curated/wind.csv,sha256=77_YrP5LttXjuwahndUiPagjwQB9Uh2tpk3Xa8-JFY4,4482
-britekit/install/yaml/base_config.yaml,sha256=vQ1xtglVkFW2qUo8ls8llpk_1YjdfSdxBJSXPmIYlUU,2744
+britekit/install/yaml/base_config.yaml,sha256=CDgTJWzIL1qSO-prJ9sMFQe1naVbXEEPDVy41VQqXOo,2765
 britekit/install/yaml/samples/cfg_infer.yaml,sha256=GsBqkL5xPlS1V-bZmXNXGFDS1_hBEGkuer8sqO3CiQw,65
 britekit/install/yaml/samples/train_dla.yaml,sha256=D2xHIgcPMOsoK2doPsF8PmoAVtkl0v2ojuuKfzYv8XI,228
 britekit/install/yaml/samples/train_effnet.yaml,sha256=MD6LEL94FN3Hd05j8N3CHn88LIOp-sERdbyWBSKfk4g,323
@@ -85,11 +85,11 @@ britekit/commands/_search.py,sha256=b7cIFSI3AuPhreYGO8HGqX12TKaOJ5uoc1uiqid04Mk,
 britekit/commands/_train.py,sha256=40Zdu9FZkmdA0bIuTK-oADw-4X6aMQYS1XGigheocdo,4152
 britekit/commands/_tune.py,sha256=CFEY08SsPF_2s_fknUHJLEsho6vL_it8NoOhB2RoVU0,7468
 britekit/commands/_wav2mp3.py,sha256=2Q4cjT6OhJmBPTNzGRMrDd6dSdBBufuQdjhH1V8ghLo,2167
-britekit/commands/_xeno.py,sha256=_6YxQ7xFdaSy5DNUaigkbYp3E8EhtOhTC9b6OFS0MFA,6026
+britekit/commands/_xeno.py,sha256=4qS2iXeGMl0jYvTAvhSvX0VvIlp7STlX27o273WwBHk,6165
 britekit/commands/_youtube.py,sha256=_u1LrwY_2GxllKd505N_2ArFMbACQ_PtVxuqUCYxFe0,2214
 britekit/core/__init__.py,sha256=QcjcFyvO5KqJLF_HBeqiCk925uU5jTUjIV5lJix9XY4,556
-britekit-0.0.11.dist-info/METADATA,sha256=d1VgnfrMq3-tnSABq7aXN8SAMsuZI5ifbPM9MB7yDtE,18553
-britekit-0.0.11.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-britekit-0.0.11.dist-info/entry_points.txt,sha256=ycnPy5DLX14RTf7lKfkQAVyIf1B1zTL1gMsHm455wmg,46
-britekit-0.0.11.dist-info/licenses/LICENSE.txt,sha256=kPoHm6iop8-CUa_720Tt8gqyvLD6D_7218u1hCCpErk,1092
-britekit-0.0.11.dist-info/RECORD,,
+britekit-0.0.12.dist-info/METADATA,sha256=-D6Wr6aoJZ7Vy23qV4_dgxSdEBD1ZqZ-ObGVDPLFmsk,18553
+britekit-0.0.12.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+britekit-0.0.12.dist-info/entry_points.txt,sha256=ycnPy5DLX14RTf7lKfkQAVyIf1B1zTL1gMsHm455wmg,46
+britekit-0.0.12.dist-info/licenses/LICENSE.txt,sha256=kPoHm6iop8-CUa_720Tt8gqyvLD6D_7218u1hCCpErk,1092
+britekit-0.0.12.dist-info/RECORD,,