README.md

metadata

library_name: transformers
language:
  - ru
license: apache-2.0
base_model: PekingU/rtdetr_r50vd_coco_o365
tags:
  - generated_from_trainer
model-index:
  - name: RT-DETR Russian car plate detection with classification by type
    results: []

RT-DETR Russian car plate detection with classification by type

This model is a fine-tuned version of PekingU/rtdetr_r50vd_coco_o365 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.1673
• Map: 0.8829
• Map 50: 0.9858
• Map 75: 0.9736
• Map Car-plates-and-these-types: -1.0
• Map Large: 0.9689
• Map Medium: 0.9125
• Map N P: 0.857
• Map P P: 0.9087
• Map Small: 0.696
• Mar 1: 0.8686
• Mar 10: 0.9299
• Mar 100: 0.9357
• Mar 100 Car-plates-and-these-types: -1.0
• Mar 100 N P: 0.9169
• Mar 100 P P: 0.9545
• Mar Large: 0.9844
• Mar Medium: 0.958
• Mar Small: 0.8354

Model description

Модель детекции номерных знаков автомобилей РФ, в данный момент 2 класса n_p и p_p, обычные номера и полицейские

Intended uses & limitations

Пример использования:

from transformers import AutoModelForObjectDetection, AutoImageProcessor
import torch
import supervision as sv


DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = AutoModelForObjectDetection.from_pretrained('Garon16/rtdetr_r50vd_russia_plate_detector').to(DEVICE)
processor = AutoImageProcessor.from_pretrained('Garon16/rtdetr_r50vd_russia_plate_detector')

path = 'path/to/image'
image = Image.open(path)
inputs = processor(image, return_tensors="pt").to(DEVICE)
with torch.no_grad():
    outputs = model(**inputs)
w, h = image.size
results = processor.post_process_object_detection(
    outputs, target_sizes=[(h, w)], threshold=0.3)
detections = sv.Detections.from_transformers(results[0]).with_nms(0.3)
labels = [
    model.config.id2label[class_id]
    for class_id
    in detections.class_id
]

annotated_image = image.copy()
annotated_image = sv.BoundingBoxAnnotator().annotate(annotated_image, detections)
annotated_image = sv.LabelAnnotator().annotate(annotated_image, detections, labels=labels)
  
grid = sv.create_tiles(
  [annotated_image],
  grid_size=(1, 1),
  single_tile_size=(512, 512),
  tile_padding_color=sv.Color.WHITE,
  tile_margin_color=sv.Color.WHITE
)
sv.plot_image(grid, size=(10, 10))

Training and evaluation data

Обучал на своём датасете - https://universe.roboflow.com/testcarplate/russian-license-plates-classification-by-this-type

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch 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: 300
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map 50	Map 75	Map Car-plates-and-these-types	Map Large	Map Medium	Map N P	Map P P	Map Small	Mar 1	Mar 10	Mar 100	Mar 100 Car-plates-and-these-types	Mar 100 N P	Mar 100 P P	Mar Large	Mar Medium	Mar Small
No log	1.0	109	64.6127	0.035	0.0558	0.0379	-1.0	0.0039	0.0663	0.0191	0.0508	0.0071	0.1523	0.3009	0.3361	-1.0	0.3179	0.3543	0.7625	0.3788	0.1157
No log	2.0	218	15.4008	0.8237	0.9418	0.9327	-1.0	0.893	0.879	0.7945	0.8529	0.4319	0.8203	0.8924	0.9018	-1.0	0.8766	0.9269	0.9656	0.9324	0.7653
No log	3.0	327	9.4050	0.8439	0.9566	0.9479	-1.0	0.9439	0.8908	0.8158	0.872	0.5171	0.8416	0.908	0.9144	-1.0	0.9002	0.9286	0.9781	0.9368	0.8051
No log	4.0	436	7.9164	0.8493	0.9665	0.9543	-1.0	0.9567	0.8903	0.8338	0.8648	0.5581	0.8481	0.9159	0.9267	-1.0	0.9173	0.936	0.975	0.949	0.8185
70.2867	5.0	545	6.8177	0.8525	0.9723	0.9602	-1.0	0.9521	0.8918	0.8234	0.8816	0.6025	0.8438	0.9214	0.9279	-1.0	0.9181	0.9378	0.975	0.9492	0.8211
70.2867	6.0	654	6.0182	0.854	0.9744	0.9619	-1.0	0.9574	0.8912	0.8251	0.8829	0.6123	0.8438	0.9176	0.927	-1.0	0.9137	0.9403	0.9781	0.9503	0.8163
70.2867	7.0	763	5.4024	0.8731	0.9772	0.9667	-1.0	0.9635	0.9113	0.8462	0.9001	0.6376	0.8608	0.9275	0.9336	-1.0	0.9202	0.9471	0.9781	0.956	0.8266
70.2867	8.0	872	5.2224	0.8726	0.9809	0.9767	-1.0	0.9582	0.9069	0.8487	0.8966	0.6472	0.8625	0.9265	0.9301	-1.0	0.9137	0.9464	0.9875	0.9528	0.8232
70.2867	9.0	981	4.7844	0.8679	0.9821	0.9687	-1.0	0.9574	0.9023	0.8451	0.8907	0.6382	0.8606	0.9213	0.9283	-1.0	0.9119	0.9448	0.9844	0.952	0.8165
4.2466	10.0	1090	5.1437	0.8729	0.9816	0.9762	-1.0	0.9577	0.9028	0.8448	0.901	0.6686	0.8605	0.9296	0.9359	-1.0	0.9203	0.9514	0.9781	0.9567	0.8413
4.2466	11.0	1199	4.5169	0.8858	0.9828	0.9768	-1.0	0.9707	0.9162	0.8628	0.9087	0.6734	0.8695	0.9264	0.931	-1.0	0.9121	0.95	0.9781	0.9538	0.823
4.2466	12.0	1308	4.5858	0.8813	0.9865	0.9744	-1.0	0.9623	0.9126	0.8585	0.9041	0.6815	0.8671	0.9308	0.9355	-1.0	0.9185	0.9526	0.9812	0.9583	0.8308
4.2466	13.0	1417	4.5345	0.8778	0.9843	0.9726	-1.0	0.957	0.9101	0.8526	0.903	0.6754	0.8628	0.9281	0.9335	-1.0	0.9158	0.9512	0.9812	0.9557	0.8314
3.589	14.0	1526	4.3003	0.8885	0.9857	0.9759	-1.0	0.9656	0.9189	0.8642	0.9128	0.6957	0.8724	0.9334	0.9375	-1.0	0.9194	0.9555	0.9875	0.959	0.8375
3.589	15.0	1635	4.3999	0.8819	0.986	0.9741	-1.0	0.9606	0.9118	0.8575	0.9064	0.6892	0.8659	0.9283	0.9336	-1.0	0.9137	0.9534	0.9844	0.9566	0.8245
3.589	16.0	1744	4.2719	0.8796	0.986	0.9726	-1.0	0.9661	0.9093	0.8543	0.905	0.6914	0.8649	0.927	0.9313	-1.0	0.9121	0.9505	0.9875	0.9543	0.8266
3.589	17.0	1853	4.2497	0.8838	0.9845	0.9733	-1.0	0.9656	0.9141	0.8599	0.9077	0.6997	0.8678	0.9295	0.9352	-1.0	0.9141	0.9562	0.9812	0.958	0.832
3.589	18.0	1962	4.2807	0.8829	0.9855	0.9754	-1.0	0.9673	0.9121	0.8558	0.9099	0.6964	0.8683	0.9286	0.9337	-1.0	0.9126	0.9548	0.9844	0.9555	0.8357
3.2442	19.0	2071	4.1978	0.8835	0.9861	0.9748	-1.0	0.9675	0.9121	0.8559	0.911	0.6932	0.8691	0.9272	0.9336	-1.0	0.9134	0.9538	0.9844	0.9557	0.8337
3.2442	20.0	2180	4.1673	0.8829	0.9858	0.9736	-1.0	0.9689	0.9125	0.857	0.9087	0.696	0.8686	0.9299	0.9357	-1.0	0.9169	0.9545	0.9844	0.958	0.8354

Framework versions

• Transformers 4.46.0.dev0
• Pytorch 2.5.0+cu124
• Tokenizers 0.20.1