ploppy06 commited on
Commit
97030f5
1 Parent(s): 60ea6fd

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +88 -72
README.md CHANGED
@@ -70,11 +70,22 @@ dataset_summary: '
70
 
71
  <!-- Provide a quick summary of the dataset. -->
72
 
 
73
 
 
74
 
 
75
 
76
- ![image/png](dataset_preview.png)
 
 
 
 
77
 
 
 
 
 
78
 
79
  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 24984 samples.
80
 
@@ -107,52 +118,56 @@ session = fo.launch_app(dataset)
107
 
108
  <!-- Provide a longer summary of what this dataset is. -->
109
 
 
 
 
 
110
 
111
-
112
- - **Curated by:** [More Information Needed]
113
- - **Funded by [optional]:** [More Information Needed]
114
- - **Shared by [optional]:** [More Information Needed]
115
- - **Language(s) (NLP):** en
116
  - **License:** bsd-2-clause
117
 
118
- ### Dataset Sources [optional]
119
 
120
  <!-- Provide the basic links for the dataset. -->
121
 
122
- - **Repository:** [More Information Needed]
123
- - **Paper [optional]:** [More Information Needed]
124
- - **Demo [optional]:** [More Information Needed]
125
 
126
  ## Uses
127
 
128
  <!-- Address questions around how the dataset is intended to be used. -->
129
 
130
- ### Direct Use
131
-
132
- <!-- This section describes suitable use cases for the dataset. -->
133
-
134
- [More Information Needed]
135
-
136
- ### Out-of-Scope Use
137
-
138
- <!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
139
 
140
- [More Information Needed]
 
141
 
142
  ## Dataset Structure
143
 
144
  <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
145
 
146
- [More Information Needed]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
147
 
148
  ## Dataset Creation
149
 
150
- ### Curation Rationale
151
-
152
- <!-- Motivation for the creation of this dataset. -->
153
-
154
- [More Information Needed]
155
-
156
  ### Source Data
157
 
158
  <!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
@@ -161,74 +176,75 @@ session = fo.launch_app(dataset)
161
 
162
  <!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
163
 
164
- [More Information Needed]
165
 
166
  #### Who are the source data producers?
167
 
168
  <!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
169
 
170
- [More Information Needed]
171
 
172
- ### Annotations [optional]
173
 
174
  <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
175
 
176
- #### Annotation process
177
-
178
- <!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
179
-
180
- [More Information Needed]
181
 
182
- #### Who are the annotators?
183
 
184
- <!-- This section describes the people or systems who created the annotations. -->
185
-
186
- [More Information Needed]
187
-
188
- #### Personal and Sensitive Information
189
-
190
- <!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
191
-
192
- [More Information Needed]
193
-
194
- ## Bias, Risks, and Limitations
195
-
196
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
 
 
 
 
 
 
 
 
 
 
 
197
 
198
- [More Information Needed]
199
 
200
- ### Recommendations
201
 
202
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
203
 
204
- Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
205
 
206
- ## Citation [optional]
207
 
208
  <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
209
 
210
  **BibTeX:**
 
 
 
 
 
 
 
 
 
 
211
 
212
- [More Information Needed]
213
-
214
- **APA:**
215
-
216
- [More Information Needed]
217
-
218
- ## Glossary [optional]
219
-
220
- <!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
221
-
222
- [More Information Needed]
223
-
224
- ## More Information [optional]
225
 
226
- [More Information Needed]
227
 
228
- ## Dataset Card Authors [optional]
229
 
230
- [More Information Needed]
231
 
232
- ## Dataset Card Contact
233
 
234
- [More Information Needed]
 
70
 
71
  <!-- Provide a quick summary of the dataset. -->
72
 
73
+ MPII Human Pose dataset is a state of the art benchmark for evaluation of articulated human pose estimation.
74
 
75
+ The dataset includes around **25K images** containing over **40K people with annotated body joints**.
76
 
77
+ The images were systematically collected using an established **taxonomy of every day human activities**.
78
 
79
+ Overall the dataset covers **410 human activities** and each image is provided with an activity label.
80
+
81
+ Each image was extracted from a **YouTube video** and provided with preceding and following un-annotated frames.
82
+
83
+ In addition, for the test set, richer annotations were obtained including body part occlusions and 3D torso and head orientations.
84
 
85
+ Following the best practices for the performance evaluation benchmarks in the literature we withhold the test annotations to prevent overfitting and tuning on the test set.
86
+ We are working on an automatic evaluation server and performance analysis tools based on rich test set annotations.
87
+
88
+ ![image/png](dataset_preview.png)
89
 
90
  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 24984 samples.
91
 
 
118
 
119
  <!-- Provide a longer summary of what this dataset is. -->
120
 
121
+ >MPII Human Pose Dataset, Version 1.0
122
+ Copyright 2015 Max Planck Institute for Informatics
123
+ Licensed under the Simplified BSD License
124
+ Annotations and the corresponding are freely available for research purposes. Commercial use is not allowed due to the fact that the authors do not have the copyright for the images themselves.
125
 
 
 
 
 
 
126
  - **License:** bsd-2-clause
127
 
128
+ ### Dataset Sources
129
 
130
  <!-- Provide the basic links for the dataset. -->
131
 
132
+ - **Website:** http://human-pose.mpi-inf.mpg.de/#
133
+ - **Paper:** http://human-pose.mpi-inf.mpg.de/contents/andriluka14cvpr.pdf
 
134
 
135
  ## Uses
136
 
137
  <!-- Address questions around how the dataset is intended to be used. -->
138
 
139
+ At the time when dataset was released **(2014)**, dataset was evaluated on 2 main tasks:
 
 
 
 
 
 
 
 
140
 
141
+ * Multi-Person Pose Estimation
142
+ * Single Person Pose Estimation
143
 
144
  ## Dataset Structure
145
 
146
  <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
147
 
148
+ ```
149
+ Name: MPII Human Pose
150
+ Media type: image
151
+ Num samples: 24984
152
+ Persistent: True
153
+ Tags: ['version1', 'MPII Human Pose']
154
+ Sample fields:
155
+ id: fiftyone.core.fields.ObjectIdField
156
+ filepath: fiftyone.core.fields.StringField
157
+ tags: fiftyone.core.fields.ListField(fiftyone.core.fields.StringField)
158
+ metadata: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.metadata.ImageMetadata)
159
+ rectangle_id: fiftyone.core.fields.ListField(fiftyone.core.fields.IntField)
160
+ activity: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.labels.Classifications)
161
+ head_rect: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.labels.Detections)
162
+ objpos: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.labels.Keypoints)
163
+ scale: fiftyone.core.fields.VectorField
164
+ annopoints: fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.labels.Keypoints)
165
+ video_id: fiftyone.core.fields.StringField
166
+ frame_sec: fiftyone.core.fields.IntField
167
+ ```
168
 
169
  ## Dataset Creation
170
 
 
 
 
 
 
 
171
  ### Source Data
172
 
173
  <!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
 
176
 
177
  <!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
178
 
179
+ See **Section 2. Dataset - Data collection paragraph** of this [paper](http://human-pose.mpi-inf.mpg.de/contents/andriluka14cvpr.pdf)
180
 
181
  #### Who are the source data producers?
182
 
183
  <!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
184
 
185
+ [Source data producers](http://human-pose.mpi-inf.mpg.de/#contact)
186
 
187
+ ### Annotations
188
 
189
  <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
190
 
191
+ Annotation description
 
 
 
 
192
 
193
+ Annotations are stored in a matlab structure **RELEASE** having following fields
194
 
195
+ ```
196
+ .annolist(imgidx) - annotations for image imgidx
197
+ .image.name - image filename
198
+ .annorect(ridx) - body annotations for a person ridx
199
+ .x1, .y1, .x2, .y2 - coordinates of the head rectangle
200
+ .scale - person scale w.r.t. 200 px height
201
+ .objpos - rough human position in the image
202
+ .annopoints.point - person-centric body joint annotations
203
+ .x, .y - coordinates of a joint
204
+ id - joint id (0 - r ankle, 1 - r knee, 2 - r hip, 3 - l hip, 4 - l knee, 5 - l ankle, 6 - pelvis, 7 - thorax, 8 - upper neck, 9 - head top, 10 - r wrist, 11 - r elbow, 12 - r shoulder, 13 - l shoulder, 14 - l elbow, 15 - l wrist)
205
+ is_visible - joint visibility
206
+ .vidx - video index in video_list
207
+ .frame_sec - image position in video, in seconds
208
+
209
+ img_train(imgidx) - training/testing image assignment
210
+
211
+ single_person(imgidx) - contains rectangle id ridx of sufficiently separated individuals
212
+
213
+ act(imgidx) - activity/category label for image imgidx
214
+ act_name - activity name
215
+ cat_name - category name
216
+ act_id - activity id
217
+
218
+ video_list(videoidx) - specifies video id as is provided by YouTube. To watch video on youtube go to https://www.youtube.com/watch?v=video_list(videoidx)
219
 
220
+ ```
221
 
222
+ #### Annotation process
223
 
224
+ <!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
225
 
226
+ See **Section 2. Dataset - Data annotation paragraph** of this [paper](http://human-pose.mpi-inf.mpg.de/contents/andriluka14cvpr.pdf)
227
 
228
+ ## Citation
229
 
230
  <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
231
 
232
  **BibTeX:**
233
+ ```
234
+ @inproceedings{andriluka14cvpr,
235
+ author = {Mykhaylo Andriluka and Leonid Pishchulin and Peter Gehler and Schiele, Bernt}
236
+ title = {2D Human Pose Estimation: New Benchmark and State of the Art Analysis},
237
+ booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
238
+ year = {2014},
239
+ month = {June}
240
+ }
241
+ ```
242
+ ## More Information
243
 
 
 
 
 
 
 
 
 
 
 
 
 
 
244
 
245
+ The following [**Github repo**](https://github.com/loloMD/51_contribution/tree/mpii_human_pose/mpii_human_pose) contains code to parse the raw data (in MATLAB format) and convert it into a FiftyOne Dataset
246
 
 
247
 
248
+ “Dataset conversion and data card contributed by [Loic Mandine](https://lolomd.github.io/)
249
 
 
250