All projects

Tiny-human detection for aerial rescue

A controlled ablation showing which architecture changes actually help find people smaller than eight pixels in disaster imagery.

Status
Thesis completed
Role
Designed the ablation, trained and evaluated four detector variants, rejected ineffective complexity, and prepared the deployable model and drone-footage pipeline.
Computer vision thesis2026
  1. 01Aerial frame
  2. 02P2 + CBAM
  3. 03Detection
  4. 04Rescue review
High-level flow only. Open the case study for decisions, evidence, and limitations.

People in aerial disaster scenes are often only a few pixels wide, occluded by debris, and surrounded by textures that defeat standard detectors.

  • Identified the stride-4 P2 detection scale as the useful change, raising very-tiny recall from .743 to .757.
  • Rejected a Mamba neck after it added 2.4M parameters and about 2.8× latency without improving accuracy.
  • Raised very-tiny recall to .850 with 1280px test-time augmentation when a higher-latency operating mode is acceptable.

Ablate, don't decorate

Starting from YOLO11m, the study added CBAM, a P2 head, and a bidirectional state-space neck one at a time under one protocol. The contribution is as much the negative result as the winning model.

The useful change

The P2 scale operates at stride four, preserving spatial detail for sub-eight-pixel targets. It improved AP50 and very-tiny recall for about half a million parameters and one millisecond of latency.

Limits kept visible

Results use one semi-synthetic dataset and one seed per configuration. Desktop-GPU latency is not an airborne-device measurement; validation on real rescue imagery remains future work.

Next projectCelloscope