DETR ResNet-50 Finetuned 10K CPPE5 Open-Source Object Detection Model - Accurately Identify Personal Protective Equipment

Detr Resnet 50 Finetuned 10k Cppe5

Developed by qubvel-hf

An object detection model fine-tuned on the CPPE-5 dataset based on facebook/detr-resnet-50, specializing in personal protective equipment detection.

Object Detection

Transformers

Open Source License:Apache-2.0 #Medical Protective Equipment Detection #Small Object Detection Optimization #End-to-End Object Detection

Downloads 4,074

Release Time : 4/18/2024

Model Overview

This model is a ResNet-50 variant of the DETR (Detection Transformer) architecture, specifically fine-tuned for medical and personal protective equipment (PPE) detection tasks.

Model Features

End-to-End Object Detection

Uses Transformer architecture for end-to-end object detection, eliminating the need for anchor box designs in traditional methods.

Personal Protective Equipment Detection

Optimized specifically for PPE items in medical settings, such as protective suits, face shields, gloves, goggles, and masks.

Medium-Scale Model

Balanced design based on ResNet-50, achieving a good trade-off between accuracy and computational efficiency.

Model Capabilities

Object detection in images

Personal protective equipment recognition

Multi-category object detection

Use Cases

Medical Safety

PPE Compliance Check

Automatically detects whether medical staff are wearing protective equipment correctly.

Protective suit detection accuracy 65.8%, face shield detection accuracy 36.9%

Industrial Safety

Workplace Safety Monitoring

Monitors whether workers in factories or construction sites are wearing necessary protective equipment.

🚀 detr-resnet-50-finetuned-10k-cppe5

This model is a fine - tuned version of [facebook/detr - resnet - 50](https://huggingface.co/facebook/detr - resnet - 50) on the cppe - 5 dataset. It is designed for object detection tasks and can achieve relatively good performance in detecting objects in the cppe - 5 dataset.

🚀 Quick Start

This model is mainly used for object detection. You can load and use it through the Hugging Face Transformers library.

✨ Features

Fine - tuned: Based on the pre - trained model facebook/detr - resnet - 50, it is fine - tuned on the cppe - 5 dataset to better adapt to the object detection tasks of this dataset.
Multiple evaluation metrics: It provides a variety of evaluation metrics, such as Loss, Map, Map 50, etc., to comprehensively evaluate the performance of the model.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

No code examples are provided in the original document.

📚 Documentation

Model Performance

It achieves the following results on the evaluation set:

Loss: 0.9865
Map: 0.3578
Map 50: 0.6781
Map 75: 0.3105
Map Small: 0.3578
Map Medium: - 1.0
Map Large: - 1.0
Mar 1: 0.365
Mar 10: 0.535
Mar 100: 0.5483
Mar Small: 0.5483
Mar Medium: - 1.0
Mar Large: - 1.0
Map Coverall: 0.6584
Mar 100 Coverall: 0.7772
Map Face Shield: 0.3691
Mar 100 Face Shield: 0.6063
Map Gloves: 0.2477
Mar 100 Gloves: 0.4266
Map Goggles: 0.1766
Mar 100 Goggles: 0.4655
Map Mask: 0.3371
Mar 100 Mask: 0.4661

Training and Evaluation Data

The model is trained on the cppe - 5 dataset. However, more detailed information about the data is not provided in the original document.

Training Procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 1337
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 100.0
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 Coverall	Mar 100 Coverall	Map Face Shield	Mar 100 Face Shield	Map Gloves	Mar 100 Gloves	Map Goggles	Mar 100 Goggles	Map Mask	Mar 100 Mask
3.7198	1.0	107	3.1869	0.0036	0.0144	0.0006	0.0036	-1.0	-1.0	0.017	0.0426	0.0647	0.0647	-1.0	-1.0	0.0169	0.1994	0.0	0.0	0.0001	0.0177	0.0	0.0	0.0008	0.1062
3.0393	2.0	214	2.8978	0.007	0.021	0.003	0.007	-1.0	-1.0	0.0192	0.0796	0.1238	0.1238	-1.0	-1.0	0.0333	0.5216	0.0	0.0	0.0001	0.0292	0.0	0.0	0.0017	0.0684
2.7812	3.0	321	2.5445	0.0138	0.0414	0.0085	0.0138	-1.0	-1.0	0.0307	0.0994	0.1258	0.1258	-1.0	-1.0	0.0655	0.4938	0.0	0.0	0.0002	0.0354	0.0	0.0	0.0033	0.1
2.5992	4.0	428	2.3828	0.0232	0.0601	0.0155	0.0232	-1.0	-1.0	0.0423	0.1202	0.1518	0.1518	-1.0	-1.0	0.1021	0.5481	0.0	0.0	0.0006	0.0495	0.0059	0.0109	0.0072	0.1503
2.3828	5.0	535	2.2672	0.0283	0.0703	0.0179	0.0283	-1.0	-1.0	0.0521	0.1283	0.1737	0.1737	-1.0	-1.0	0.1344	0.5846	0.0	0.0	0.001	0.0833	0.0	0.0	0.0063	0.2006
2.2633	6.0	642	2.0618	0.0479	0.0996	0.0416	0.0479	-1.0	-1.0	0.0782	0.1679	0.2035	0.2035	-1.0	-1.0	0.2099	0.6333	0.003	0.0159	0.0018	0.1187	0.0052	0.0218	0.0195	0.2277
2.1837	7.0	749	2.1100	0.0455	0.1159	0.0255	0.0455	-1.0	-1.0	0.0747	0.1582	0.1894	0.1894	-1.0	-1.0	0.2068	0.6185	0.0085	0.0556	0.001	0.0734	0.0002	0.0018	0.0113	0.1977
2.0689	8.0	856	2.0000	0.054	0.1389	0.0301	0.054	-1.0	-1.0	0.0954	0.1846	0.2159	0.2159	-1.0	-1.0	0.2155	0.5537	0.0314	0.1397	0.0049	0.1406	0.0002	0.0018	0.0181	0.2435
2.0417	9.0	963	1.8702	0.0697	0.1631	0.0501	0.0697	-1.0	-1.0	0.1074	0.2173	0.257	0.257	-1.0	-1.0	0.2826	0.6086	0.0279	0.181	0.0046	0.1734	0.0102	0.0418	0.0234	0.2802
1.9972	10.0	1070	1.8563	0.0742	0.1568	0.0541	0.0742	-1.0	-1.0	0.1196	0.2416	0.2786	0.2786	-1.0	-1.0	0.2933	0.6086	0.0233	0.1921	0.0053	0.1672	0.0239	0.0891	0.025	0.3362
1.8931	11.0	1177	1.6778	0.1054	0.2248	0.0898	0.1054	-1.0	-1.0	0.1456	0.2764	0.3033	0.3033	-1.0	-1.0	0.3955	0.671	0.0498	0.2603	0.0108	0.2188	0.0149	0.0382	0.056	0.3282
1.8269	12.0	1284	1.6905	0.1111	0.2399	0.0942	0.1111	-1.0	-1.0	0.1543	0.2949	0.3257	0.3257	-1.0	-1.0	0.4113	0.679	0.069	0.319	0.0087	0.2021	0.015	0.0909	0.0514	0.3373
1.8036	13.0	1391	1.6406	0.1149	0.2407	0.097	0.1149	-1.0	-1.0	0.1636	0.3108	0.3372	0.3372	-1.0	-1.0	0.4255	0.6759	0.0771	0.3381	0.0109	0.2182	0.0137	0.1309	0.047	0.3226
1.7463	14.0	1498	1.7169	0.1106	0.2421	0.0875	0.1106	-1.0	-1.0	0.1776	0.3205	0.3511	0.3511	-1.0	-1.0	0.3996	0.7	0.0404	0.2476	0.0117	0.2458	0.0257	0.2036	0.0757	0.3582
1.763	15.0	1605	1.5961	0.1245	0.2577	0.1018	0.1245	-1.0	-1.0	0.1817	0.3384	0.3677	0.3677	-1.0	-1.0	0.4575	0.6679	0.0775	0.3698	0.0107	0.2505	0.0318	0.1964	0.0447	0.3537
1.6467	16.0	1712	1.5365	0.1376	0.3073	0.1062	0.1376	-1.0	-1.0	0.2164	0.38	0.408	0.408	-1.0	-1.0	0.455	0.6852	0.0739	0.3873	0.0215	0.2719	0.0442	0.2891	0.0934	0.4068
1.6222	17.0	1819	1.5990	0.1295	0.2696	0.1026	0.1295	-1.0	-1.0	0.1802	0.3409	0.3693	0.3693	-1.0	-1.0	0.4577	0.6654	0.0786	0.3619	0.0297	0.2958	0.0211	0.2218	0.0603	0.3017
1.6239	18.0	1926	1.4164	0.159	0.3543	0.1262	0.159	-1.0	-1.0	0.235	0.3929	0.4138	0.4138	-1.0	-1.0	0.4753	0.7204	0.0921	0.3968	0.039	0.2922	0.0323	0.2636	0.1565	0.396
1.5448	19.0	2033	1.4689	0.1628	0.3725	0.1314	0.1628	-1.0	-1.0	0.205	0.3811	0.4064	0.4064	-1.0	-1.0	0.4794	0.6895	0.1038	0.419	0.0398	0.2828	0.0333	0.28	0.1578	0.3605
1.5026	20.0	2140	1.4093	0.1798	0.397	0.1369	0.1798	-1.0	-1.0	0.2336	0.4125	0.4349	0.4349	-1.0	-1.0	0.4851	0.6858	0.1494	0.4508	0.0341	0.2859	0.0434	0.3382	0.1869	0.4136
1.4797	21.0	2247	1.4605	0.1652	0.3605	0.1254	0.1652	-1.0	-1.0	0.2295	0.3823	0.4041	0.4041	-1.0	-1.0	0.4978	0.6957	0.0968	0.3825	0.0529	0.2797	0.0263	0.3236	0.1522	0.339
1.4298	22.0	2354	1.4231	0.163	0.3558	0.115	0.163	-1.0	-1.0	0.2256	0.3851	0.4108	0.4108	-1.0	-1.0	0.4902	0.7093	0.1033	0.4159	0.0515	0.313	0.0261	0.3109	0.1437	0.3051
1.4157	23.0	2461	1.3665	0.1914	0.4048	0.1533	0.1914	-1.0	-1.0	0.2478	0.4232	0.447	0.447	-1.0	-1.0	0.491	0.6975	0.1599	0.4683	0.0502	0.3021	0.0603	0.3618	0.1956	0.4051
1.4438	24.0	2568	1.2908	0.2103	0.433	0.168	0.2103	-1.0	-1.0	0.2643	0.4512	0.4761	0.4761	-1.0	-1.0	0.5368	0.7136	0.1493	0.4873	0.0789	0.3609	0.043	0.3891	0.2433	0.4294
1.4044	25.0	2675	1.4752	0.1709	0.3749	0.1388	0.1709	-1.0	-1.0	0.2187	0.3926	0.4191	0.4191	-1.0	-1.0	0.4862	0.7167	0.09	0.3905	0.0762	0.299	0.0393	0.3527	0.1627	0.3367
1.3703	26.0	2782	1.3047	0.2162	0.4568	0.1714	0.2162	-1.0	-1.0	0.2661	0.4344	0.4548	0.4548	-1.0	-1.0	0.5342	0.7272	0.166	0.4508	0.0971	0.3281	0.0424	0.3527	0.2414	0.4153
1.3292	27.0	2889	1.2674	0.22	0.4681	0.1702	0.22	-1.0	-1.0	0.2743	0.4286	0.4473	0.4473	-1.0	-1.0	0.5438	0.7265	0.2128	0.4429	0.1171	0.3443	0.0387	0.3455	0.1878	0.3774
1.359	28.0	2996	1.3156	0.2007	0.4272	0.1536	0.2007	-1.0	-1.0	0.2715	0.4384	0.4555	0.4555	-1.0	-1.0	0.5306	0.7111	0.163	0.5016	0.0896	0.3135	0.0307	0.38	0.1898	0.3712
1.3471	29.0	3103	1.2646	0.2161	0.455	0.172	0.2161	-1.0	-1.0	0.277	0.4492	0.4728	0.4728	-1.0	-1.0	0.5301	0.7216	0.1708	0.519	0.1216	0.3271	0.0391	0.4145	0.2189	0.3819
1.308	30.0	3210	1.3017	0.2107	0.4465	0.1718	0.2107	-1.0	-1.0	0.2556	0.4141	0.4387	0.4387	-1.0	-1.0	0.5321	0.7136	0.1531	0.454	0.1037	0.3203	0.0334	0.3218	0.2313	0.3836
1.3023	31.0	3317	1.2809	0.2174	0.462	0.1714	0.2174	-1.0	-1.0	0.2646	0.4242	0.4473	0.4473	-1.0	-1.0	0.5484	0.7259	0.1686	0.427	0.1163	0.3536	0.0506	0.3564	0.2029	0.3734
1.2561	32.0	3424	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...

🔧 Technical Details

The model is based on the DETR (Detection Transformer) architecture, which uses a Transformer - based encoder - decoder structure for object detection. During training, the Adam optimizer is used, and a linear learning rate scheduler is adopted. The mixed - precision training technique (Native AMP) is also applied to speed up the training process.

📄 License

This model is released under the Apache 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご