Oxford IIIT Pet Dataset (fixed)

Dataset Description

The Oxford‑IIIT Pet Dataset is a collection of 37 pet categories, each with roughly 200 images, created by the Oxford Visual Geometry Group. Images exhibit large variations in scale, pose, and illumination. All images are provided with ground‑truth annotations for breed, head region of interest, and pixel‑level three‑class segmentation.

Repaired Images

The original dataset contained several corrupted images, which have been repaired and uploaded to this repository. The following are the repaired images and their restoration steps:

1. Abyssinian_34.jpg

   • File header: GIF image data, version 89a, 250 × 202
   • Converted to JPEG format.

2. Egyptian_Mau_139.jpg

   • File header: GIF image data, version 89a, 350 × 250
   • Converted to JPEG format.

3. Egyptian_Mau_145.jpg

   • File header: GIF image data, version 89a, 216 × 188
   • Converted to JPEG format.

4. Egyptian_Mau_167.jpg

   • File header: GIF image data, version 89a, 183 × 27
   • Converted to JPEG format.

5. Egyptian_Mau_177.jpg

   • File header: GIF image data, version 87a, 300 × 214
   • Converted to JPEG format.

6. beagle_116.jpg

   • Corrupted JPEG data: data segment ends prematurely
   • Loaded with OpenCV and saved again as JPEG. The corrupted region could not be easily repaired.

7. chihuahua_121.jpg

   • Corrupted JPEG data: 240 extra bytes before marker 0xd9
   • Loaded with OpenCV and saved again as JPEG. This is a simple method to clean unnecessary bytes from the file.