rtdetr-v2-setup8

This model is a fine-tuned version of PekingU/rtdetr_v2_r50vd on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 6.3398
• Map: 0.6331
• Map50: 0.9685
• Map75: 0.7375
• Mar 100: 0.6865
• Accuracy: 0.9517

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map50	Map75	Mar 100	Accuracy
811.1910	1.0	36	420.2512	0.0	0.0	0.0	0.0	0.0
411.5360	2.0	72	75.6202	0.0	0.0	0.0	0.0	0.0
60.4656	3.0	108	22.9026	0.5109	0.7942	0.6183	0.5638	0.7651
29.3075	4.0	144	9.9040	0.5747	0.896	0.699	0.6355	0.8467
14.2247	5.0	180	7.6852	0.6105	0.9701	0.7129	0.6745	0.9205
13.0624	6.0	216	6.8710	0.5906	0.9644	0.6442	0.6582	0.9195
12.3203	7.0	252	6.6475	0.6033	0.9744	0.6688	0.6652	0.9521
12.0858	8.0	288	6.5161	0.6149	0.9626	0.7359	0.6745	0.932
11.8652	9.0	324	6.3703	0.6303	0.9556	0.7491	0.6794	0.9252
11.3376	10.0	360	6.4277	0.6317	0.9612	0.7524	0.6844	0.8896
11.2978	11.0	396	6.1840	0.6333	0.9755	0.7348	0.695	0.9653
10.9792	12.0	432	6.3293	0.6266	0.958	0.7376	0.683	0.9189
10.8468	13.0	468	6.2389	0.6442	0.9661	0.7834	0.6957	0.9452
10.5343	14.0	504	6.3311	0.6299	0.9576	0.7306	0.6872	0.9444
10.3091	15.0	540	6.3600	0.621	0.9653	0.7056	0.6858	0.9133
10.3811	16.0	576	6.2518	0.6393	0.9781	0.7671	0.6943	0.9329
10.3929	17.0	612	6.2909	0.6405	0.9779	0.7622	0.6957	0.9267
10.1662	18.0	648	6.3008	0.6342	0.9678	0.7475	0.6865	0.9583
10.0634	19.0	684	6.2901	0.6358	0.9678	0.7464	0.6894	0.9583
10.1193	20.0	720	6.3398	0.6331	0.9685	0.7375	0.6865	0.9517

Framework versions

• Transformers 5.6.2
• Pytorch 2.5.1+cu124
• Datasets 4.8.4
• Tokenizers 0.22.2