jrziehn commited on
Commit
9da580b
1 Parent(s): 23c591c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +42 -4
README.md CHANGED
@@ -43,7 +43,8 @@ of optical/imaging effects on the classification ability.
43
 
44
 
45
  - **Curated by:** [Anne Sielemann](https://www.linkedin.com/in/anne-sielemann-23011026a/), [Stefan Wolf](https://www.linkedin.com/in/stefan-wolf-2552211a9/),
46
- - [Jens Ziehn](https://www.linkedin.com/in/jrziehn/), Masoud Roschani, and Juergen Beyerer. [Fraunhofer IOSB](https://www.iosb.fraunhofer.de/), Germany.
 
47
  - **Funded by:**
48
  - [Fraunhofer](https://www.fraunhofer.de/en.html) Internal Programs under Grant No. PREPARE 40-02702 within the ML4Safety project
49
  - the [Ministry of Economic Affairs, Labour and Housing](https://wm.baden-wuerttemberg.de/) of the state of [Baden-Wuerttemberg](https://www.thelaend.de/), Germany, as part of the FeinSyn research project
@@ -81,7 +82,11 @@ The dataset is intended for the following use cases:
81
 
82
  <!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
83
 
84
- [More Information Needed]
 
 
 
 
85
 
86
  ## Dataset Structure
87
 
@@ -204,7 +209,37 @@ The dataset contains no data that might be considered personal, sensitive, or pr
204
 
205
  <!-- This section is meant to convey both technical and sociotechnical limitations. -->
206
 
207
- [More Information Needed]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
208
 
209
  ### Recommendations
210
 
@@ -212,7 +247,10 @@ The dataset contains no data that might be considered personal, sensitive, or pr
212
 
213
  <!-- Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations. -->
214
 
215
- ## Citation [optional]
 
 
 
216
 
217
  <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
218
 
 
43
 
44
 
45
  - **Curated by:** [Anne Sielemann](https://www.linkedin.com/in/anne-sielemann-23011026a/), [Stefan Wolf](https://www.linkedin.com/in/stefan-wolf-2552211a9/),
46
+ [Jens Ziehn](https://www.linkedin.com/in/jrziehn/), Masoud Roschani, and Juergen Beyerer. [Fraunhofer IOSB](https://www.iosb.fraunhofer.de/), Germany.
47
+
48
  - **Funded by:**
49
  - [Fraunhofer](https://www.fraunhofer.de/en.html) Internal Programs under Grant No. PREPARE 40-02702 within the ML4Safety project
50
  - the [Ministry of Economic Affairs, Labour and Housing](https://wm.baden-wuerttemberg.de/) of the state of [Baden-Wuerttemberg](https://www.thelaend.de/), Germany, as part of the FeinSyn research project
 
82
 
83
  <!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
84
 
85
+ The dataset should not be used for critical applications, particularly high-risk applications as named by the European AI Act under Annex III (which includes "AI systems intended to be used for the ‘real-time’ and ‘post’ remote biometric
86
+ identification of natural persons" and "AI systems intended to be used as safety components in the management and operation of road traffic"), without exhaustive
87
+ research into the fitness of the dataset, to evaluate whether it is "relevant, sufficiently representative, and to the best extent possible free of errors and complete
88
+ in view of the intended purpose of the system." No such claim is not made with the publication of this dataset.
89
+
90
 
91
  ## Dataset Structure
92
 
 
209
 
210
  <!-- This section is meant to convey both technical and sociotechnical limitations. -->
211
 
212
+
213
+ - **Model variants:** Every class in the dataset contains only a single variant, while in practice, optional or retrofitted equipment such as
214
+ sunroofs or fog lights may vary within one make/model/year class. The accuracy of the 3D models is not evaluated separately.
215
+ Different national variants of the vehicles are not reflected in the dataset. The dataset covers European, Asian and American models,
216
+ but has a strong focus on models and variants common in Western Europe.
217
+
218
+ - **Vehicle lights:** The 3D models were not annotated for individual vehicle light functions, such that no distinction between
219
+ daytime lights, high beam, turn indicators, etc., is made in the dataset. The lights are not triggered individually.
220
+
221
+ - **License plates:** License plates are modeled as part of
222
+ the 3D mesh and textures and are therefore fixed for each
223
+ vehicle geometry. Some vehicles feature fixed license plate
224
+ numbers or logos, others contain empty license plates or no
225
+ license plates. Therefore, in the dataset without masked license
226
+ plates, license plate appearance will be identical across cars of the same class, and shared among some different classes as well.
227
+
228
+ - **Environment:** Environment variation is limited to over-
229
+ all lighting conditions and road model and textures. No com-
230
+ plex shadows or reflections from roadside objects, other ve-
231
+ hicles, occlusions, environment conditions (snow, raindrops,
232
+ fog, ...) or low light / nighttime conditions are included.
233
+
234
+ - **Environment and lighting:** Available light models are currently limited
235
+ and not calibrated. Therefore, no absolute scales are given
236
+ and the relative vehicle light brightness (and corresponding
237
+ effects) will be incorrect. Surface properties for physically-based rendering
238
+ are selected qualitatively and are not based on accurate physical measurements.
239
+
240
+ - **Perspective and camera:** Only frontal perspective images are included in the dataset, and only one set of intrinsic
241
+ camera parameters is used, and only a single camera lens type (based on a Tamron M112FM35 35 mm lens) and only a very limited set of imaging artifacts are simulated.
242
+
243
 
244
  ### Recommendations
245
 
 
247
 
248
  <!-- Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations. -->
249
 
250
+ It is recommended to use the dataset primarily for scientific research. Application to practical real-world use cases should include
251
+ human oversight and the exhaustive evaluation of the fitness for the respective purpose, including the impact of domain shifts.
252
+
253
+ ## Citation
254
 
255
  <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
256