PaGE ViT B Distill Base distilled student; human level performance. Part of the PaGE gaze target estimation family. Backbone: DINOv3 ViT B (derivative of DINOv3, full parameter trained) Params: ~90M Scene input: 512×512, Head input: 256×256, Heatmap output: 64×64 Source checkpoint: vitb distill.pt Self contained weights This checkpoint includes the full DINOv3 backbone weights in its safetensors files. No external DINOv3 weights are downloaded — the DINOv3 model structure is provided by transformers==5.6.2 (built in dinov3 vit ), and the backbone weights here are derivative weights from full parameter training of DINOv3. The model code ( modeling page.py ) is loaded automatically from Octopus1/PaGE via auto map when you pass trust remote code=True . Installation Tested with transformers 5.6.2. Usage Inputs / Outputs See the family README for the full spec. Input dict: images (list of [B,3,512,512] ), head images (list of [sum(Np),3,256,256] ), bboxes (per image list of (xmin,ymin,xmax,ymax) in [0,1] ). Output dict: heatmap (list of [Np,64,64] , sigmoid), inout (list of [Np] , sigmoid). License The PaGE decoder and gaze heads are released under the MIT License (see LICENSE ). The DI…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy