DogReID-1553: A Large-Scale Dog Re-Identification Video Dataset

Dataset Description This dataset is developed for dog re-identification (Re-ID) across videos and images. It contains 7,463 videos of 1,553 distinct dog identities (averaging 4.8 videos/dog). Bounding boxes are manually annotated on the first frame of each video to support localization. Benchmark, Leaderboard & Baseline To facilitate ongoing research and provide a standard starting point, we have established a public benchmark alongside this dataset. We provide a baseline model implementation, and we host an active leaderboard to track state-of-the-art methods. Leaderboard: Submit your results and view current rankings. Baseline Implementation: Check out our GitHub repository for the starter code and Re-ID pipeline. Paper: Detailed methodology and baseline evaluation (under review). Data Collection To avoid the restrictive licensing of scraped social media data, videos were crowdsourced directly from users via our submission website, community promotion (Reddit, Discord, Instagram), and the Prolific research platform. All submitters explicitly agreed to release their videos under the permissive Creative Commons CC0 1.0 license. Pre-processing and Anonymization Videos were converted to MP4 (H.264), stripped of audio, and temporally cropped so the target dog is clearly visible in the first frame. To ensure privacy, all videos were manually reviewed. We used video editing software to permanently black out sensitive information (faces, license plates, tattoos, addresses) from the dataset. Data Leakage Prevention To prevent models from overfitting to environmental cues (e.g., memorizing a specific backyard instead of the dog), the gallery and query sets are scene-disjoint. Videos for each identity were manually partitioned to ensure zero background overlap between query and gallery splits. Directory Structure Files are organized by unique UUIDs (DOG_ID and VIDEO_ID): /Videos/ — Contains the .mp4 files. /Videos/DOG_ID/DOG_ID-VIDEO_ID.mp4 Example: /Videos/00ab4db8-3e94-44e3-a4b1-9fe316c67b60/00ab4db8-3e94-44e3-a4b1-9fe316c67b60-2cc7b112-2af8-4466-931a-c83a2466d81a.mp4 /Images/ — Contains the .jpg first-frame crops. /Images/DOG_ID/DOG_ID-VIDEO_ID.jpg Annotations and Metadata bounding_boxes.csv — Top-left coordinates and dimensions (in pixels) for the first-frame bounding boxes. DOG_ID,VIDEO_ID,x_top_left,y_top_left,width,height 0062cada-a402-41bb-980e-6ae5e0672440,3f13a35c-be97-4331-b1f9-503bf4628e00,3,259,459,503 breeds.csv — User-submitted dog breeds. DOG_ID,VIDEO_ID,BREED 44833f04-7e94-4e7b-9549-4a0b19cd4f5f,e16c3a9f-df3b-46da-b211-e04515e0df44,Pitlabmix splits.csv — Defines the 50-50 identity splits for evaluation. The GROUP column enforces the scene-disjoint subsets. DOG_ID,VIDEO_ID,GROUP,SPLIT_CLOSED_SET,SPLIT_OPEN_SET 0062cada-a402-41bb-980e-6ae5e0672440,4f82f8da-1e3f-46f8-b48a-3a161d52cdf0,0,gallery,query Evaluation Splits Summary The dataset is divided into two primary evaluation settings. For both, Videos are shown first, followed by the number of unique Identities in parentheses. 1. Closed-world Setting Goal: Identify a dog from a set where every query identity is guaranteed to exist in the gallery. Train Set: 3,788 Videos (776 IDs) Gallery Set: 1,998 Videos (777 IDs) . Query Set: 1,677 Videos (680 IDs) 2. Open-world Setting Goal: Identify a dog while also determining if the dog even exists in the gallery at all. Train Set: 3,708 Videos (766 IDs) Gallery Set: 1,296 Videos (512 IDs) Query Set: 2,459 Videos (777 IDs) File Formats Videos: .mp4 Images: .jpg Metadata, splits, and annotations: .csv