biofeatures-segmentation-model-v0

This model is a fine-tuned version of facebook/dinov2-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0008
• Mean Iou: 0.1121
• Mean Accuracy: 0.5494
• Overall Accuracy: 0.4350
• Per Category Iou: [0.428875787306029, 0.0, 0.019716575276221136, 0.0]
• Per Category Accuracy: [0.4310967895930302, nan, 0.6676298940449884, nan]

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Per Category Iou	Per Category Accuracy
No log	1.0	1	1.4059	0.1427	0.4695	0.5579	[0.5562818599789431, 0.0, 0.014499935879462083, 0.0]	[0.5609604964792935, nan, 0.3779817553402459, nan]
No log	2.0	3	1.3840	0.1411	0.4723	0.5513	[0.5494584994796633, 0.0, 0.014742997450507332, 0.0]	[0.5540078768349445, nan, 0.3905603716924472, nan]
No log	3.0	5	1.2995	0.1347	0.4875	0.5256	[0.5228650852904727, 0.0, 0.015928569127170795, 0.0]	[0.5268972431077694, nan, 0.44818403308969346, nan]
No log	4.0	6	1.2613	0.1318	0.4959	0.5140	[0.510847356110949, 0.0, 0.016524158390199304, 0.0]	[0.5146580737558181, nan, 0.47719417530738284, nan]
No log	5.0	7	1.2262	0.1292	0.5028	0.5033	[0.49970093057513276, 0.0, 0.016994310950270652, 0.0]	[0.5033097028284998, nan, 0.5023514080117853, nan]
No log	6.0	9	1.1640	0.1245	0.5162	0.4846	[0.4803372678082809, 0.0, 0.017850364304430988, 0.0]	[0.48353836973385844, nan, 0.5488696243413225, nan]
0.705	7.0	11	1.1132	0.1207	0.5273	0.4693	[0.46447068022188864, 0.0, 0.01851722825766549, 0.0]	[0.46734741615944625, nan, 0.5872853986061534, nan]
0.705	8.0	12	1.0918	0.1192	0.5315	0.4630	[0.4578840273256509, 0.0, 0.018755083066793923, 0.0]	[0.4606325337152405, nan, 0.6023570740551872, nan]
0.705	9.0	13	1.0728	0.1177	0.5357	0.4571	[0.45179794244230087, 0.0, 0.018991934358516368, 0.0]	[0.45443227115407564, nan, 0.6169188055980509, nan]
0.705	10.0	15	1.0414	0.1154	0.5422	0.4478	[0.44210904125769945, 0.0, 0.019351865020141773, 0.0]	[0.4445513784461153, nan, 0.6399229418097343, nan]
0.705	11.0	17	1.0189	0.1136	0.5467	0.4409	[0.4349064013303538, 0.0, 0.019587747425958247, 0.0]	[0.43722258026017424, nan, 0.6562411468071846, nan]
0.705	12.0	18	1.0110	0.1130	0.5482	0.4385	[0.4324789458237841, 0.0, 0.019663744413421932, 0.0]	[0.4347516410072801, nan, 0.6616238880389824, nan]
0.705	13.0	19	1.0049	0.1124	0.5487	0.4362	[0.43008841824517946, 0.0, 0.01968208824323486, 0.0]	[0.4323313044516052, nan, 0.6650235140801178, nan]
0.5715	13.33	20	1.0008	0.1121	0.5494	0.4350	[0.428875787306029, 0.0, 0.019716575276221136, 0.0]	[0.4310967895930302, nan, 0.6676298940449884, nan]

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.0
• Datasets 2.18.0
• Tokenizers 0.15.2