PASCAL Scene/Object Means for dataset trainval (original) (raw)
Image and Bounding Box Means for PASCAL Visual Object Classes Challenge 2006: Dataset trainval
Generated by Tomasz Malisiewicz on Sat Feb 25 2006
A summary of The PASCAL Visual Object Classes Challenge 2006
What did I do?
For each image that contains a particular object, I averaged the images by resizing each image to 400 x 400, summing in RGB space (default in MATLAB), dividing by the number of images, then resizing to the mean size of the images containing that object.
I repeated that procedure for the bounding boxes that enclose those objects. I simply resized the bounding box of each object (per pose) to 100 x 100, summed, divided by the number of bounding boxes for that object and that particular pose, then resized to the mean bounding box size.
What are these images of?
The first column is the average of the entire image. Note that on the average, images containing cows are larger than images containing dogs. The second column averages all of the bounding boxes, while the remaining columns are averages of the bounding boxes per pose. The final column averages the 'unspecified' pose bounding boxes. Note how frontal bounding boxes of cars are significantly smaller than Right or Left bounding boxes of cars (there are also less of them). This is probably because it is harder to take an image of a car's front compared to its side.
Also, one can view the summary montage of the above images (each image is 100 x 100).
Next to each image is a number which denotes the number of images averaged.
| | Image Containing Object | [Bounding Box for All Poses] | [Frontal BB] | [Rear BB] | [Left BB] | [Right BB] | [Unspecified BB] | |
| ------------------------- | ----------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| bicycle | 
276 | 
323 | 
11 | 
6 | 
111 | 
58 | 
137 |
| bus | 
177 | 
235 | 
16 | 
13 | 
15 | 
12 | 
179 |
| car | 
576 | 
854 | 
59 | 
149 | 
144 | 
117 | 
385 |
| cat | 
387 | 
429 | 
43 | 
1 | 
16 | 
15 | 
354 |
| cow | 
207 | 
313 | 
23 | 
4 | 
76 | 
62 | 
148 |
| dog | 
369 | 
422 | 
44 | 
6 | 
36 | 
31 | 
305 |
| horse | 
248 | 
326 | 
38 | 
8 | 
63 | 
52 | 
165 |
| motorbike | 
239 | 
275 | 
16 | 
5 | 
32 | 
54 | 
168 |
| person | 
733 | 
1156 | 
261 | 
91 | 
88 | 
78 | 
638 |
| sheep | 
251 | 
421 | 
29 | 
8 | 
98 | 
91 | 
195 |
Also, one can view the summary montage of the above images (each image is 100 x 100).
Generated by Tomasz Malisiewicz