Files changed (1) hide show
  1. README.md +101 -58
README.md CHANGED
@@ -1,8 +1,27 @@
1
  ---
2
  library_name: transformers
3
- tags: []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  ---
5
 
 
6
  # Model Card for Model ID
7
 
8
  <!-- Provide a quick summary of what the model is/does. -->
@@ -13,24 +32,41 @@ tags: []
13
 
14
  ### Model Description
15
 
16
- <!-- Provide a longer summary of what this model is. -->
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
 
18
  This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
19
 
20
- - **Developed by:** [More Information Needed]
21
- - **Funded by [optional]:** [More Information Needed]
22
- - **Shared by [optional]:** [More Information Needed]
23
- - **Model type:** [More Information Needed]
24
- - **Language(s) (NLP):** [More Information Needed]
25
- - **License:** [More Information Needed]
26
- - **Finetuned from model [optional]:** [More Information Needed]
 
 
27
 
28
  ### Model Sources [optional]
29
 
30
  <!-- Provide the basic links for the model. -->
31
 
32
- - **Repository:** [More Information Needed]
33
- - **Paper [optional]:** [More Information Needed]
34
  - **Demo [optional]:** [More Information Needed]
35
 
36
  ## Uses
@@ -41,26 +77,20 @@ This is the model card of a 🤗 transformers model that has been pushed on the
41
 
42
  <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
 
44
- [More Information Needed]
45
 
46
  ### Downstream Use [optional]
47
 
48
  <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
 
50
- [More Information Needed]
51
-
52
  ### Out-of-Scope Use
53
 
54
  <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
 
56
- [More Information Needed]
57
-
58
  ## Bias, Risks, and Limitations
59
 
60
  <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
 
62
- [More Information Needed]
63
-
64
  ### Recommendations
65
 
66
  <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
@@ -71,7 +101,32 @@ Users (both direct and downstream) should be made aware of the risks, biases and
71
 
72
  Use the code below to get started with the model.
73
 
74
- [More Information Needed]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
75
 
76
  ## Training Details
77
 
@@ -79,55 +134,55 @@ Use the code below to get started with the model.
79
 
80
  <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
 
82
- [More Information Needed]
83
 
84
  ### Training Procedure
85
 
86
  <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
 
88
- #### Preprocessing [optional]
 
 
 
 
 
89
 
90
- [More Information Needed]
91
 
 
92
 
93
  #### Training Hyperparameters
94
 
95
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
 
 
96
 
97
  #### Speeds, Sizes, Times [optional]
98
 
99
  <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
 
101
- [More Information Needed]
102
-
103
  ## Evaluation
104
 
105
  <!-- This section describes the evaluation protocols and provides the results. -->
106
 
 
 
107
  ### Testing Data, Factors & Metrics
108
 
109
  #### Testing Data
110
 
111
  <!-- This should link to a Dataset Card if possible. -->
112
 
113
- [More Information Needed]
114
-
115
  #### Factors
116
 
117
  <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
 
119
- [More Information Needed]
120
-
121
  #### Metrics
122
 
123
  <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
 
125
- [More Information Needed]
126
-
127
  ### Results
128
 
129
- [More Information Needed]
130
-
131
  #### Summary
132
 
133
 
@@ -136,64 +191,52 @@ Use the code below to get started with the model.
136
 
137
  <!-- Relevant interpretability work for the model goes here -->
138
 
139
- [More Information Needed]
140
-
141
  ## Environmental Impact
142
 
143
  <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
 
145
  Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
 
147
- - **Hardware Type:** [More Information Needed]
148
- - **Hours used:** [More Information Needed]
149
- - **Cloud Provider:** [More Information Needed]
150
- - **Compute Region:** [More Information Needed]
151
- - **Carbon Emitted:** [More Information Needed]
152
 
153
  ## Technical Specifications [optional]
154
 
155
  ### Model Architecture and Objective
156
 
157
- [More Information Needed]
158
 
159
  ### Compute Infrastructure
160
 
161
- [More Information Needed]
162
-
163
  #### Hardware
164
 
165
- [More Information Needed]
166
-
167
  #### Software
168
 
169
- [More Information Needed]
170
-
171
  ## Citation [optional]
172
 
173
  <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
 
175
  **BibTeX:**
176
 
177
- [More Information Needed]
 
 
 
 
 
 
 
 
 
178
 
179
  **APA:**
180
 
181
- [More Information Needed]
182
-
183
  ## Glossary [optional]
184
 
185
  <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
 
187
- [More Information Needed]
188
-
189
  ## More Information [optional]
190
 
191
- [More Information Needed]
192
-
193
  ## Model Card Authors [optional]
194
 
195
- [More Information Needed]
196
-
197
- ## Model Card Contact
198
 
199
- [More Information Needed]
 
1
  ---
2
  library_name: transformers
3
+ license: apache-2.0
4
+ language:
5
+ - en
6
+ pipeline_tag: object-detection
7
+ tags:
8
+ - object-detection
9
+ - vision
10
+ datasets:
11
+ - coco
12
+ widget:
13
+ - src: >-
14
+ https://huggingface.co/datasets/mishig/sample_images/resolve/main/savanna.jpg
15
+ example_title: Savanna
16
+ - src: >-
17
+ https://huggingface.co/datasets/mishig/sample_images/resolve/main/football-match.jpg
18
+ example_title: Football Match
19
+ - src: >-
20
+ https://huggingface.co/datasets/mishig/sample_images/resolve/main/airport.jpg
21
+ example_title: Airport
22
  ---
23
 
24
+
25
  # Model Card for Model ID
26
 
27
  <!-- Provide a quick summary of what the model is/does. -->
 
32
 
33
  ### Model Description
34
 
35
+ The YOLO series has become the most popular framework for real-time object detection due to its reasonable trade-off between speed and accuracy.
36
+ However, we observe that the speed and accuracy of YOLOs are negatively affected by the NMS.
37
+ Recently, end-to-end Transformer-based detectors (DETRs) have provided an alternative to eliminating NMS.
38
+ Nevertheless, the high computational cost limits their practicality and hinders them from fully exploiting the advantage of excluding NMS.
39
+ In this paper, we propose the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to our best knowledge that addresses the above dilemma.
40
+ We build RT-DETR in two steps, drawing on the advanced DETR:
41
+ first we focus on maintaining accuracy while improving speed, followed by maintaining speed while improving accuracy.
42
+ Specifically, we design an efficient hybrid encoder to expeditiously process multi-scale features by decoupling intra-scale interaction and cross-scale fusion to improve speed.
43
+ Then, we propose the uncertainty-minimal query selection to provide high-quality initial queries to the decoder, thereby improving accuracy.
44
+ In addition, RT-DETR supports flexible speed tuning by adjusting the number of decoder layers to adapt to various scenarios without retraining.
45
+ Our RT-DETR-R50 / R101 achieves 53.1% / 54.3% AP on COCO and 108 / 74 FPS on T4 GPU, outperforming previously advanced YOLOs in both speed and accuracy.
46
+ We also develop scaled RT-DETRs that outperform the lighter YOLO detectors (S and M models).
47
+ Furthermore, RT-DETR-R50 outperforms DINO-R50 by 2.2% AP in accuracy and about 21 times in FPS.
48
+ After pre-training with Objects365, RT-DETR-R50 / R101 achieves 55.3% / 56.2% AP. The project page: this [https URL](https://zhao-yian.github.io/RTDETR/).
49
+
50
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6579e0eaa9e58aec614e9d97/WULSDLsCVs7RNEs9KB0Lr.png)
51
 
52
  This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
53
 
54
+ - **Developed by:** Yian Zhao and Sangbum Choi
55
+ - **Funded by [optional]:** National Key R&D Program of China (No.2022ZD0118201), Natural Science Foundation of China (No.61972217, 32071459, 62176249, 62006133, 62271465),
56
+ and the Shenzhen Medical Research Funds in China (No.
57
+ B2302037).
58
+ - **Shared by [optional]:** Sangbum Choi
59
+ - **Model type:**
60
+ - **Language(s) (NLP):**
61
+ - **License:** Apache-2.0
62
+ - **Finetuned from model [optional]:**
63
 
64
  ### Model Sources [optional]
65
 
66
  <!-- Provide the basic links for the model. -->
67
 
68
+ - **Repository:** https://github.com/lyuwenyu/RT-DETR
69
+ - **Paper [optional]:** https://arxiv.org/abs/2304.08069
70
  - **Demo [optional]:** [More Information Needed]
71
 
72
  ## Uses
 
77
 
78
  <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
79
 
80
+ You can use the raw model for object detection. See the [model hub](https://huggingface.co/models?search=rtdetr) to look for all available RTDETR models.
81
 
82
  ### Downstream Use [optional]
83
 
84
  <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
85
 
 
 
86
  ### Out-of-Scope Use
87
 
88
  <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
89
 
 
 
90
  ## Bias, Risks, and Limitations
91
 
92
  <!-- This section is meant to convey both technical and sociotechnical limitations. -->
93
 
 
 
94
  ### Recommendations
95
 
96
  <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
 
101
 
102
  Use the code below to get started with the model.
103
 
104
+ ```
105
+ import torch
106
+ import requests
107
+
108
+ from PIL import Image
109
+ from transformers import RTDetrForObjectDetection, RTDetrImageProcessor
110
+
111
+ url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
112
+ image = Image.open(requests.get(url, stream=True).raw)
113
+
114
+ image_processor = RTDetrImageProcessor.from_pretrained("PekingU/rtdetr_r101vd")
115
+ model = RTDetrForObjectDetection.from_pretrained("PekingU/rtdetr_r101vd")
116
+
117
+ inputs = image_processor(images=image, return_tensors="pt")
118
+
119
+ with torch.no_grad():
120
+ outputs = model(**inputs)
121
+
122
+ results = image_processor.post_process_object_detection(outputs, target_sizes=torch.tensor([image.size[::-1]]), threshold=0.3)
123
+
124
+ for result in results:
125
+ for score, label_id, box in zip(result["scores"], result["labels"], result["boxes"]):
126
+ score, label = score.item(), label_id.item()
127
+ box = [round(i, 2) for i in box.tolist()]
128
+ print(f"{model.config.id2label[label]}: {score:.2f} {box}")
129
+ ```
130
 
131
  ## Training Details
132
 
 
134
 
135
  <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
136
 
137
+ The RTDETR model was trained on [COCO 2017 object detection](https://cocodataset.org/#download), a dataset consisting of 118k/5k annotated images for training/validation respectively.
138
 
139
  ### Training Procedure
140
 
141
  <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
142
 
143
+ We conduct experiments on
144
+ COCO [20] and Objects365 [35], where RT-DETR is trained
145
+ on COCO train2017 and validated on COCO val2017
146
+ dataset. We report the standard COCO metrics, including
147
+ AP (averaged over uniformly sampled IoU thresholds ranging from 0.50-0.95 with a step size of 0.05), AP50, AP75, as
148
+ well as AP at different scales: APS, APM, APL.
149
 
150
+ #### Preprocessing [optional]
151
 
152
+ Images are resized/rescaled such that the shortest side is at 640 pixels.
153
 
154
  #### Training Hyperparameters
155
 
156
+ - **Training regime:** <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
157
+
158
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6579e0eaa9e58aec614e9d97/E15I9MwZCtwNIms-W8Ra9.png)
159
 
160
  #### Speeds, Sizes, Times [optional]
161
 
162
  <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
163
 
 
 
164
  ## Evaluation
165
 
166
  <!-- This section describes the evaluation protocols and provides the results. -->
167
 
168
+ This model achieves an AP (average precision) of 53.1 on COCO 2017 validation. For more details regarding evaluation results, we refer to table 2 of the original paper.
169
+
170
  ### Testing Data, Factors & Metrics
171
 
172
  #### Testing Data
173
 
174
  <!-- This should link to a Dataset Card if possible. -->
175
 
 
 
176
  #### Factors
177
 
178
  <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
179
 
 
 
180
  #### Metrics
181
 
182
  <!-- These are the evaluation metrics being used, ideally with a description of why. -->
183
 
 
 
184
  ### Results
185
 
 
 
186
  #### Summary
187
 
188
 
 
191
 
192
  <!-- Relevant interpretability work for the model goes here -->
193
 
 
 
194
  ## Environmental Impact
195
 
196
  <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
197
 
198
  Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
199
 
 
 
 
 
 
200
 
201
  ## Technical Specifications [optional]
202
 
203
  ### Model Architecture and Objective
204
 
205
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6579e0eaa9e58aec614e9d97/sdIwTRlHNwPzyBNwHja60.png)
206
 
207
  ### Compute Infrastructure
208
 
 
 
209
  #### Hardware
210
 
 
 
211
  #### Software
212
 
 
 
213
  ## Citation [optional]
214
 
215
  <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
216
 
217
  **BibTeX:**
218
 
219
+ ```bibtex
220
+ @misc{lv2023detrs,
221
+ title={DETRs Beat YOLOs on Real-time Object Detection},
222
+ author={Yian Zhao and Wenyu Lv and Shangliang Xu and Jinman Wei and Guanzhong Wang and Qingqing Dang and Yi Liu and Jie Chen},
223
+ year={2023},
224
+ eprint={2304.08069},
225
+ archivePrefix={arXiv},
226
+ primaryClass={cs.CV}
227
+ }
228
+ ```
229
 
230
  **APA:**
231
 
 
 
232
  ## Glossary [optional]
233
 
234
  <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
235
 
 
 
236
  ## More Information [optional]
237
 
 
 
238
  ## Model Card Authors [optional]
239
 
240
+ [Sangbum Choi](https://huggingface.co/danelcsb)
 
 
241
 
242
+ ## Model Card Contact