scout-eccv-y-clip-v17

scout-eccv-y-clip-v17

Part of SCOUT (Sim-to-real Contrastive Outlier Understanding via Text), an ECCV 2026 workshop paper on JEPA-based text-based person re-identification for the AI City Challenge Track 4 (Sim2Real). Paper: SCOUT: Sim-to-Real Text-Based Person Retrieval by Embedding-Space Prediction over Frozen Video Features Code: https://github.com/abtraore/SCOUT-ECCV

v17 recipe with the frozen Y-encoder swapped from EmbeddingGemma-300M to a CLIP-ViT-L/14 text tower. The best frozen-Y-encoder result in the paper's Y-encoder ablation.

• Val mAP@10: 0.8932
• Val Hit@10: 0.9908
• Trainable / total parameters: 549.6M / 1281.9M

LP-FT stage-1 initialization for x_explora. A fusion member of the paper's headline board system.

What's in this repo

Only the parameters SCOUT actually trained: a model.safetensors (trainable predictor weights, and for ExPLoRA/LoRA variants, the adapted encoder deltas) plus a config.json recording the exact recipe. Frozen backbone weights (V-JEPA2 from Meta, the text encoder from its own source) are not included here; they're downloaded from their own public sources at load time.

Usage

from scripts.load_from_hf import load_from_hf

model = load_from_hf("Abdrah/scout-eccv-y-clip-v17").eval()
# model.encode_image(pixel_values)  # [B, 3, 384, 384] in [0, 1] -> [B, out_dim] L2-normalized
# model.encode_text(["a caption"])  # -> [B, out_dim] L2-normalized

Or drop it straight into the repo's own evaluation script:

uv run python scripts/local_eval.py --hf-repo Abdrah/scout-eccv-y-clip-v17 --pab-root /path/to/PAB --query-source val

License

Apache 2.0 (see the source repo's LICENSE/NOTICE). This checkpoint does not redistribute any frozen backbone weights, which carry their own upstream licenses.