TernScope: Ternary Model-Hardware Co-Design Cuts Accelerator Area by 88.46% for Endoscopic Segmentation
Monash · hf · 2026-11-02
TernScope co-designs a compact ternary-weight segmentation network with simplified hardware for edge endoscopic imaging. All 25 convolutions use ternary weights and 8-bit activations with multiplier-free computation, yielding a 0.2483M-param / 0.0949-GMACs model that retains competitive accuracy. Ternary Progressive QAT adapts the fused network from deep to shallow layers, cutting compute-array area by 88.46% versus the original Gemini 45nm systolic array. Code is open-sourced on GitHub.
More from Research
- New theory paper proves linear decay is the unique minimax-optimal stepsize schedule for SGD — konstmish · 2026-11-02
- Karpathy's nanochat: the best ChatGPT $100 can buy, near 60k GitHub stars — karpathy · 2026-11-02
- After reading 46 resources on agent evals, one author explains why her 95% score was misleading — blaizedsouza · 2026-11-02
- How to Loop Your Transformer: Meta/MIT arXiv paper on loop scaling laws — rohanpaul_ai · 2026-11-02
- Meta & MIT paper: looping transformer layers beats adding depth at fixed size — rohanpaul_ai · 2026-11-02
- AI tell-tale word 'delve' declining in latest NeurIPS papers, analysis shows — JungWooHa2 · 2026-11-02