Add Team_GAILforce zero-shot submission

[donghun0621]

Submission Metadata

Please fill out every field below. These values are used for benchmark review and leaderboard display.

method_name: Qwen-VLM Coarse-to-Fine CCTV Accident Detector
authors: {Donghun Ryu}
affiliation: {GAIL Lab, Chung-ang University}
paper_url: N/A
code_url: N/A
contact_email: {donghun0621@cau.ac.kr}
benchmark: zero-shot
short_description: Zero-shot CCTV accident detection pipeline using Qwen VLM with hierarchical coarse-to-fine spatial-temporal scanning. Coarse stage uses 3x3 overlapping spatial regions with ordinal score (0-5) plus phase classification (score2). Top-K candidates are clustered via IoU-based region merging, then refined at 8fps and 12fps. Type is verified on peak-centered frames, and accident point is localized via 3-round 3x3 grid zoom-in.

Checklist

• I added exactly one submission CSV under submissions/<benchmark>/
• The CSV follows the required ACCIDENT submission format
• The metadata block above is complete and accurate

[github-actions]

Scoring preview for zero-shot benchmark:

• Unified score: 0.5643
• Type accuracy: 0.5673
• Temporal @ sigma=1.0: 0.5702
• Spatial @ sigma=1.0: 0.5557

If this PR is merged, the leaderboard update workflow will open a follow-up PR with the public leaderboard change.