snap-plate-instance-localization-dfine-v4

This model is a fine-tuned version of aguxez/snap-plate-localization-dfine-v1 on the aguxez/snap-plate-instance-localization-v4 dataset. It achieves the following results on the evaluation set:

• Loss: 4.1975
• Map: 0.2631
• Map 50: 0.5376
• Map 75: 0.1749
• Map Small: -1.0
• Map Medium: 0.1537
• Map Large: 0.3632
• Mar 1: 0.1375
• Mar 10: 0.3917
• Mar 100: 0.4625
• Mar Small: -1.0
• Mar Medium: 0.3727
• Mar Large: 0.5385
• Map Weight Plate: 0.2631
• Mar 100 Weight Plate: 0.4625

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: 2e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 7
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.05
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map 50	Map 75	Map Small	Map Medium	Map Large	Mar 1	Mar 10	Mar 100	Mar Small	Mar Medium	Mar Large	Map Weight Plate	Mar 100 Weight Plate
13.6666	1.0	66	4.3026	0.2542	0.5147	0.1399	-1.0	0.1212	0.3802	0.1458	0.3625	0.4458	-1.0	0.3545	0.5231	0.2542	0.4458
14.0320	2.0	132	4.2434	0.2583	0.5265	0.145	-1.0	0.1174	0.3759	0.1625	0.3542	0.4583	-1.0	0.3636	0.5385	0.2583	0.4583
14.2001	3.0	198	4.1975	0.2631	0.5376	0.1749	-1.0	0.1537	0.3632	0.1375	0.3917	0.4625	-1.0	0.3727	0.5385	0.2631	0.4625

Framework versions

• Transformers 5.14.1
• Pytorch 2.13.0+cu130
• Datasets 5.0.1
• Tokenizers 0.22.2