Face Recognition Dataset Specialist for 90-Degree / Side-Pose Face Recognition Improvement

We are looking for an experienced AI dataset researcher / computer vision dataset specialist to help us find and provide legally usable face image datasets. The required datasets should include face images from different angles, especially: * Frontal face images * 30°–45° semi-profile face images * 60°–75° angled face images * 90° side/profile face images * Same-person images from multiple angles where possible We only need help with finding and providing the datasets. Our internal computer vision team will handle cleaning, preprocessing, pose labeling, formatting, and preparation for training. ⸻ Important Note We already have public mugshot-style datasets with frontal and side/profile images. Please do not apply if your main deliverable is only public mugshot datasets. We are looking for additional datasets that can add more variety, such as: * Natural face angles * Real-world lighting conditions * Different camera qualities * Indoor and outdoor images * Non-controlled environments * Multi-angle images of the same person * Identity-labeled face datasets * Pose or yaw-labeled face datasets ⸻ What We Need From the Freelancer The freelancer should help us with: 1. Dataset Discovery Find face image datasets that include one or more of the following: * Frontal and side/profile images * Same-person multi-angle images * Multi-pose face datasets * Identity-labeled face datasets * Pose/yaw-labeled face datasets * Real-world image conditions * Different lighting, distance, and camera-quality variations ⸻ 2. Dataset Access Provide one of the following: * Direct dataset download links * Official dataset source pages * Access instructions * Application/approval steps if the dataset requires permission * Dataset mirrors only if legally allowed ⸻ 3. License Verification For every dataset, provide a clear license summary explaining whether it can be used for: * Commercial AI training * Internal AI training * Research-only use * Benchmark/evaluation only * Redistribution or non-redistribution * Required citation or attribution We do not want unclear, risky, or illegally scraped datasets. ⸻ 4. Dataset Summary Report For each recommended dataset, provide: Dataset name Official source link License type Allowed usage Number of identities Number of images Available pose angles Whether same-person multi-angle images exist Whether identity labels exist Whether pose/yaw labels exist Download size Access requirements Known limitations Recommendation: training / validation / testing / benchmark only Dataset Types We Are Looking For We are especially interested in: * Multi-pose face datasets * Non-mugshot face datasets * Identity-labeled face datasets * Datasets with frontal and profile views * Datasets with 45°, 60°, 75°, or 90° views * Natural face-recognition datasets * Datasets with yaw/pose annotations * Datasets captured in real-world or semi-real-world conditions * Datasets with lighting and camera-quality variation ⸻ Datasets We Do Not Need Please do not propose only: * Public mugshot datasets * Random scraped internet images * Social media images without clear rights * Unlabeled image folders * Datasets without license information * Datasets that cannot be legally used for AI training ⸻ Deliverables The final delivery should include: 1. A list of recommended datasets 2. Official source links for each dataset 3. License and usage-rights summary 4. Dataset access/download instructions 5. Dataset statistics where available 6. Notes on identity labels and pose labels 7. Recommendation on which datasets are most useful 8. Small sample review, if legally allowed No training preparation is required. No preprocessing is required. No cleaning is required. We only need the datasets and their legal/technical information.