191 Commits

Author	SHA1	Message	Date
mbsantiago	2d796394f6	Store anns and preds instead of evals in EvaluatorModule	2025-09-30 19:10:07 +01:00
mbsantiago	49ec1916ce	Fix API after module name change	2025-09-30 18:24:06 +01:00
mbsantiago	8727edf466	Save checkpoint on max class mAP	2025-09-30 18:20:31 +01:00
mbsantiago	2f48c58de1	api_v2	2025-09-30 13:56:25 +01:00
mbsantiago	981e37c346	Writing batch inference code	2025-09-30 13:22:03 +01:00
mbsantiago	30159d64a9	Update example config	2025-09-28 16:22:21 +01:00
mbsantiago	c9f0c5c431	Added bbox iou affinity function	2025-09-28 16:08:21 +01:00
mbsantiago	10865ee600	Re-org gallery example plots	2025-09-28 15:45:48 +01:00
mbsantiago	87ed44c8f7	Plotting reorganised	2025-09-27 23:58:06 +01:00
mbsantiago	df2abff654	Task/Metrics restructure	2025-09-26 15:23:38 +01:00
mbsantiago	d6ddc4514c	Better evaluation organisation	2025-09-25 17:48:29 +01:00
mbsantiago	4cd983a2c2	Better train cli arg names	2025-09-18 09:44:27 +01:00
mbsantiago	e65df81db2	Evaluate using Lightning too to handle device changes	2025-09-18 09:28:21 +01:00
mbsantiago	6c25787123	Logging is not just for training	2025-09-18 09:27:40 +01:00
mbsantiago	8c80402f08	Move clips and audio to dedicated module	2025-09-18 09:27:24 +01:00
mbsantiago	b81a882b58	Add metrics and plots	2025-09-17 10:30:24 +01:00
mbsantiago	957c0735d2	Starting new API	2025-09-16 19:39:30 +01:00
mbsantiago	bbb96b33a2	Config restructuring	2025-09-16 18:57:56 +01:00
mbsantiago	7d6cba5465	Restructuring	2025-09-16 13:38:38 +01:00
mbsantiago	60e922d565	Use pascal voc map computation by default	2025-09-16 10:56:37 +01:00
mbsantiago	704b28292b	Cleaning train module	2025-09-15 16:50:08 +01:00
mbsantiago	e752e96b93	Restructure eval metrics and plotting	2025-09-15 16:01:15 +01:00
mbsantiago	ec1c0ff020	Better matching module, remove generic from classification evaluations	2025-09-14 18:16:59 +01:00
mbsantiago	8628133fd7	Compute mAP	2025-09-14 10:08:51 +01:00
mbsantiago	d80377981e	Fix plotting	2025-09-14 09:38:45 +01:00
mbsantiago	ad5293e0d0	Ad FileNotFoundError to plotting	2025-09-13 20:05:42 +01:00
mbsantiago	01e7a5df25	Add ignore at ends when evaluating	2025-09-13 19:03:40 +01:00
mbsantiago	6d70140bc9	Default to normal anchor	2025-09-13 13:56:47 +01:00
mbsantiago	4fd2e84773	Fix clip missalignment in validation dataset	2025-09-11 11:09:59 +01:00
mbsantiago	e65d5a6846	Added more clipping options for validation	2025-09-10 21:09:51 +01:00
mbsantiago	615c811bb4	Add detection_class_name to targets protocol	2025-09-09 20:30:20 +01:00
mbsantiago	41b18c3f0a	Better order for checkpoints	2025-09-09 15:56:46 +01:00
mbsantiago	16a0fa7b75	Add targets to train cli	2025-09-09 15:45:00 +01:00
mbsantiago	115084fd2b	Updat lightning version	2025-09-09 15:31:40 +01:00
mbsantiago	951dc59718	Add seed option to train	2025-09-09 13:23:56 +01:00
mbsantiago	3376be06a4	Add experiment name	2025-09-09 09:02:25 +01:00
mbsantiago	cd4955d4f3	Eval	2025-09-08 22:04:30 +01:00
mbsantiago	c73984b213	Small fixes	2025-09-08 18:35:02 +01:00
mbsantiago	d8d2e5a2c2	Remove preprocessing modules	2025-09-08 18:11:58 +01:00
mbsantiago	b056d7d28d	Make sure training is still working	2025-09-08 18:03:56 +01:00
mbsantiago	95a884ea16	Update tests	2025-09-08 18:00:17 +01:00
mbsantiago	b7ae526071	Big changes in data module	2025-09-08 17:50:25 +01:00
mbsantiago	cf6d0d1ccc	Remove stale tests	2025-09-07 11:03:46 +01:00
mbsantiago	709b6355c2	torch.multiprocessing didn't work, returning to serial processing	2025-09-01 11:27:23 +01:00
mbsantiago	db2ad11743	Make matching in parallel for speedup	2025-09-01 11:19:02 +01:00
mbsantiago	e0ecc3c3d1	Clear evaluation callback after epoch ends	2025-09-01 08:56:38 +01:00
mbsantiago	71c2301c21	Independent preprocessor for generating validation plots	2025-08-31 23:04:16 +01:00
mbsantiago	d3d2a28130	Move detections array to cpu	2025-08-31 22:59:06 +01:00
mbsantiago	5b9a5a968f	Refactor eval code	2025-08-31 22:57:02 +01:00
mbsantiago	8d093c3ca2	Use custom AugmentationSequence instead of nn.Sequential	2025-08-31 19:27:15 +01:00
mbsantiago	2f4edeffff	Create separate preprocessor for the train/val datasets	2025-08-31 19:22:45 +01:00
mbsantiago	cca1d82d63	Remove blocking .to(device)	2025-08-31 19:10:06 +01:00
mbsantiago	55f473c9ca	Fix logging issue	2025-08-31 19:06:09 +01:00
mbsantiago	40f6b64611	Remove train preprocessing	2025-08-31 18:28:52 +01:00
mbsantiago	1cec332dd5	Change default train duration to 0.256 instead of 0.512	2025-08-30 14:08:00 +01:00
mbsantiago	93e89ecc46	LR Scheduler takes num of total batches	2025-08-28 08:52:11 +01:00
mbsantiago	34ef9e92a1	Make sure preprocessing is batchable	2025-08-27 23:58:38 +01:00
mbsantiago	0b5ac96fe8	Update model config	2025-08-27 23:58:07 +01:00
mbsantiago	ff754a1269	Tweaks of augmentation config	2025-08-27 18:23:38 +01:00
mbsantiago	ed76ec24b6	Plot anchor points	2025-08-27 18:13:40 +01:00
mbsantiago	d25efdad10	Fix plotting after update	2025-08-26 11:48:06 +01:00
mbsantiago	3043230d4f	Device fixing #5	2025-08-25 23:20:30 +01:00
mbsantiago	67e37227f5	Device fixing #4	2025-08-25 23:11:33 +01:00
mbsantiago	9d4a9fc35c	Device fixing #3	2025-08-25 23:08:49 +01:00
mbsantiago	d0bab60bf3	Device fixing #2	2025-08-25 23:07:09 +01:00
mbsantiago	a267db290c	Device fixing	2025-08-25 23:04:13 +01:00
mbsantiago	441ccb3382	Remove user warning from plotting function	2025-08-25 22:49:48 +01:00
mbsantiago	281c4dcb8a	Remove xr from postprocess	2025-08-25 22:46:21 +01:00
mbsantiago	cc9e47b022	Fix plotting after changes	2025-08-25 19:07:12 +01:00
mbsantiago	1f26103f42	Cleanup train preprocessing	2025-08-25 18:37:46 +01:00
mbsantiago	c80078feee	Removing stale tests	2025-08-25 17:23:27 +01:00
mbsantiago	0bb0caddea	Update augmentations	2025-08-25 17:06:17 +01:00
mbsantiago	76dda0a0e9	Fix train preprocessing	2025-08-25 14:01:31 +01:00
mbsantiago	c36ef3ecb5	Labels to torch	2025-08-25 12:43:34 +01:00
mbsantiago	667b18a54d	Preprocessing in pytorch	2025-08-25 11:41:55 +01:00
mbsantiago	61115d562c	Moved types to dedicated module	2025-08-24 10:55:48 +01:00
mbsantiago	02adc19070	Better structure for training module	2025-08-23 18:23:45 +01:00
mbsantiago	d4f249366e	Clean up the detector head	2025-08-18 16:07:51 +01:00
mbsantiago	98f83e8b34	Small fix	2025-08-18 10:25:14 +01:00
mbsantiago	9d1497b3f4	Handle case of empty preds or gt	2025-08-17 23:00:40 +01:00
mbsantiago	7af72912da	Implement previous matching strategy (greedy)	2025-08-17 22:52:00 +01:00
mbsantiago	4aea3fb2b0	Copy librosa PCEN implementation	2025-08-17 18:02:40 +01:00
mbsantiago	c1945ebdb7	Minor difference in loss fixed	2025-08-16 18:49:30 +01:00
mbsantiago	13da5d2814	Remove unnecessary extra conv layer	2025-08-16 18:32:35 +01:00
mbsantiago	78a0975864	Same issue	2025-08-14 11:07:27 +01:00
mbsantiago	c9848deebf	Fix logging issue	2025-08-14 11:05:21 +01:00
mbsantiago	a7301bcdc8	Add image logging for mlflow	2025-08-14 10:50:00 +01:00
mbsantiago	fdbb9c2b43	Go back to serial matching	2025-08-12 19:57:49 +01:00
mbsantiago	75e555ff00	Fix matching error after optimising	2025-08-12 19:49:24 +01:00
mbsantiago	c7b110feeb	Limit number of parallel processes to match predictions and annotations	2025-08-12 19:06:44 +01:00
mbsantiago	7d92ec772b	Fix number formatting in gallery plot	2025-08-12 18:49:21 +01:00
mbsantiago	0bb703f3c1	Add call type alias	2025-08-12 18:45:54 +01:00
mbsantiago	51d0a49da9	Improve performance of postprocessing code	2025-08-12 17:47:17 +01:00
mbsantiago	b997a122f1	Enable batdetect2 in the cli	2025-08-11 12:25:04 +01:00
mbsantiago	c41551b59c	Use match config for training	2025-08-11 12:02:46 +01:00
mbsantiago	8aa2d0cd11	Add evaluation config with match strategy	2025-08-11 12:02:30 +01:00
mbsantiago	65d13a32b7	Also ignore assertion errors in gallery plotting	2025-08-11 09:46:36 +01:00
mbsantiago	31a0d1bbb5	Drop duplicate bin edges	2025-08-11 08:59:12 +01:00
mbsantiago	f361420e58	Fix issue in gallery plotting	2025-08-11 08:54:37 +01:00
mbsantiago	2cc0bd59d4	Add plot limits	2025-08-11 01:37:22 +01:00