Image Processing

A standardized benchmark dataset for evaluating AI-generated image detection tools. Contains 2,038 images with balanced split between AI-generated (1,018) and real photos (1,020). Covers 17 state-of-the-art AI generators including DALL-E 3, Midjourney, Stable Diffusion, and Flux. Images span 6…

The Airborne Object Tracking (AOT) dataset is a collection of flight sequences collected onboard aerial vehicles with high-resolution cameras. To generate those sequences, two aircraft are equipped with sensors and fly planned encounters (e.g., Helicopter1 in Figure 1(a)). The trajectories are…

Contains 10K stereo pair images collected in 8 different scenarios plus 12K stereo pair images synchronized with diver body pose measurements.

Contains 9339 images with resolution of 1600×1200 where each pixel is 0.111µmX0.111µm from 15 different slides stained with AgNOR technique, having at least one label per image. Have more than sixty-three thousand annotations. The images from patients of Gynecology and Colonoscopy Outpatient Clinic…

Character Recognition in Natural Images (both English and Kannada are available)

4890 raw 18-megapixel images, each containing a SpyderCube color target in their scenes, manually labelled categories, and ground truth illumination chromaticities. The dataset was collected for illumination estimation problem.

This data set consists of 28 video sequences of driving recorded in the CARLA simulator, resulting in a total of 10767 frames. For each frame, pixel-wise semantic labels are provided. The scenes are recorded in dynamic weather and traffic conditions.

A large-scale anime image database with 3.33m+ images annotated with 99.7m+ tags; it can be useful for machine learning purposes such as image recognition and generation.

25 well-known photographic film stocks with their characteristics in machine-readable form, for film-simulation work, photography education and image metadata tooling. The file notes that availability, datasheets and emulsion characteristics change over time and that manufacturer documentation is…

The HumanEva-I dataset contains 7 calibrated video sequences (4 grayscale and 3 color) that are synchronized with 3D body poses obtained from a motion capture system. The database contains 4 subjects performing a 6 common actions (e.g. walking, jogging, gesturing, etc.). The error metrics for…

Contains over 10 million labeled images related to biology and conservation, particularly images from motion-triggered camera traps.

This 360° video dataset contains 20 stitched and unstitched 8K/30 fps stereoscopic videos, lasting from 248 to 1207 seconds.

Contains 16 driving scenarios with 6 camera views each. Manually ROI labels describe circular salient image regions in the driving context. It has about 16000 images and labels.

This dataset consists of extracted visual content for 16,358,041 historic newspaper pages in Chronicling America. The visual content was identified using an object detection model trained on annotations of World War 1-era Chronicling America pages, including annotations made by volunteers as part…

Pictures with segmentation masks for 2.8 million object instances in 350 categories.

Contains images of text written in one of two Arabic fonts (Ruqaa and Nastaliq (Farsi)). The dataset contains 40,000 synthesized image and 516 real one, 40,516 in total. Images are in RGB JPG format at 100×100px.

25.000 synthetic scenery's across ten different vehicle interiors for several simulated sensor inputs and ground truth data

Crowd Violence / Non-violence Database and benchmark

Data still available for download from Wayback

Synthetic benchmark for font family identification. Contains 11,995 JPG images of single words set in 600 known fonts (sans-serif, serif, slab serif, monospaced), rendered on real surface photos, scenes and printed objects at three difficulty levels, with the font, text, word box and per-letter…