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). 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.