vit-human-action-recognition

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 7.1743
• Accuracy: 0.0663

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.1662	0.1269	100	3.2481	0.0435
1.1397	0.2538	200	3.6602	0.0807
0.8699	0.3807	300	4.1130	0.0365
0.8009	0.5076	400	3.8621	0.0720
0.8207	0.6345	500	4.1116	0.0704
0.7429	0.7614	600	4.2491	0.0641
0.7514	0.8883	700	4.5465	0.0654
0.4544	1.0152	800	4.5093	0.0778
0.3475	1.1421	900	4.8903	0.0763
0.3919	1.2690	1000	4.9154	0.0689
0.3912	1.3959	1100	5.3429	0.0687
0.3395	1.5228	1200	5.0278	0.0789
0.4903	1.6497	1300	5.7057	0.0533
0.4362	1.7766	1400	5.6518	0.0530
0.3901	1.9036	1500	5.5715	0.0737
0.1338	2.0305	1600	5.5777	0.0876
0.202	2.1574	1700	6.2254	0.0669
0.2534	2.2843	1800	6.5145	0.0559
0.2006	2.4112	1900	6.5269	0.0544
0.1209	2.5381	2000	6.1272	0.0719
0.1793	2.6650	2100	6.4983	0.0496
0.1034	2.7919	2200	6.2938	0.0611
0.0916	2.9188	2300	6.5665	0.0657
0.0538	3.0457	2400	6.7762	0.0659
0.0596	3.1726	2500	6.8592	0.0669
0.0114	3.2995	2600	7.0222	0.06
0.0708	3.4264	2700	7.1938	0.0552
0.0097	3.5533	2800	7.1244	0.0635
0.0371	3.6802	2900	7.1808	0.0633
0.0304	3.8071	3000	7.1658	0.0669
0.0065	3.9340	3100	7.1743	0.0663

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1