detr-finetuned-cppe-5-10k-steps

This model is a fine-tuned version of facebook/detr-resnet-50 on the cppe-5 dataset. It achieves the following results on the evaluation set:

• Loss: 1.6238
• Map: 0.1619
• Map 50: 0.3226
• Map 75: 0.1429
• Map Small: 0.0501
• Map Medium: 0.129
• Map Large: 0.227
• Mar 1: 0.1773
• Mar 10: 0.3167
• Mar 100: 0.3392
• Mar Small: 0.128
• Mar Medium: 0.2626
• Mar Large: 0.4711
• Map Coverall: 0.4278
• Mar 100 Coverall: 0.6532
• Map Face Shield: 0.1078
• Mar 100 Face Shield: 0.2937
• Map Gloves: 0.0679
• Mar 100 Gloves: 0.2991
• Map Goggles: 0.0102
• Mar 100 Goggles: 0.0985
• Map Mask: 0.1959
• Mar 100 Mask: 0.3516

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: 4
• eval_batch_size: 8
• seed: 1337
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map 50	Map 75	Map Coverall	Map Face Shield	Map Gloves	Map Goggles	Map Mask	Map Large	Map Medium	Map Small	Mar 1	Mar 10	Mar 100	Mar 100 Coverall	Mar 100 Face Shield	Mar 100 Gloves	Mar 100 Goggles	Mar 100 Mask	Mar Large	Mar Medium	Mar Small
2.5499	1.0	213	2.3248	0.0367	0.0798	0.0325	0.1697	0.0	0.0078	0.0	0.0058	0.0417	0.0114	0.0018	0.0555	0.1278	0.1648	0.5104	0.0	0.1813	0.0	0.1324	0.2026	0.1044	0.0416
2.1119	2.0	426	2.0867	0.0493	0.1064	0.0379	0.1978	0.0	0.0253	0.0	0.0236	0.0574	0.0304	0.0072	0.0805	0.1667	0.2069	0.5874	0.0	0.204	0.0	0.2431	0.2325	0.1572	0.073
2.0052	3.0	639	2.1689	0.0563	0.1279	0.0441	0.0154	0.0414	0.0688	0.0818	0.1611	0.1793	0.0616	0.131	0.2199	0.2039	0.4757	0.0	0.0	0.0158	0.1942	0.0	0.0	0.0618	0.2267
1.9373	4.0	852	1.9264	0.0813	0.1755	0.0679	0.0125	0.056	0.0953	0.0916	0.1816	0.206	0.0662	0.1446	0.2549	0.3464	0.6302	0.0	0.0	0.0245	0.2004	0.0	0.0	0.0357	0.1996
1.8396	5.0	1065	1.8418	0.0958	0.208	0.0731	0.0194	0.0751	0.1163	0.116	0.2195	0.2399	0.095	0.1802	0.2939	0.3278	0.6054	0.0228	0.1127	0.031	0.2071	0.0	0.0	0.0975	0.2742
1.7659	6.0	1278	1.8737	0.1004	0.2399	0.0791	0.0377	0.0913	0.1207	0.1226	0.2204	0.2382	0.0808	0.1912	0.295	0.2906	0.5856	0.0439	0.1165	0.0497	0.2281	0.0	0.0	0.1178	0.2609
1.6415	7.0	1491	1.7200	0.1305	0.2822	0.105	0.044	0.1026	0.172	0.1505	0.2689	0.2891	0.1047	0.2239	0.386	0.3916	0.6257	0.0516	0.1962	0.0526	0.2554	0.0059	0.0523	0.1508	0.316
1.6405	8.0	1704	1.6820	0.1411	0.2866	0.1303	0.0542	0.1162	0.1854	0.1572	0.2867	0.3088	0.1204	0.2432	0.4042	0.4147	0.6347	0.0549	0.2468	0.0592	0.271	0.0053	0.0631	0.1713	0.3284
1.5513	9.0	1917	1.6380	0.1546	0.3144	0.1307	0.0569	0.1215	0.2131	0.17	0.3014	0.323	0.112	0.2563	0.4431	0.4352	0.6414	0.0804	0.2544	0.0657	0.2871	0.0062	0.0846	0.1853	0.3476
1.5564	10.0	2130	1.6238	0.1619	0.3226	0.1429	0.0501	0.129	0.227	0.1773	0.3167	0.3392	0.128	0.2626	0.4711	0.4278	0.6532	0.1078	0.2937	0.0679	0.2991	0.0102	0.0985	0.1959	0.3516

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.7.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.1