Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU
Details
- External ID
- 48527623
- Source
- HN
- Company
- —
- Product
- Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU
- Website domain
- github.com
- Launched
- June 14, 2026
- Cohort
- —
- Upvotes
- 72
- Upvotes percentile
- 0.8866120218579235
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Enrichment
- Theme
- autonomous robotics, drones, and sensor hardware
- Vertical
- Horizontal
- Function
- Hardware & robotics
- Audience
- Developer
- AI stance
- AI feature
- Project type
- Commercial product
- Normalized one-liner
- uav detection on edge hardware
- Manually corrected
- False
Could you build this?
No Running dual real-time YOLOv8 models at 42 FPS on an RK3588S requires embedded Linux optimization, Rockchip NPU (RKNN) C++ SDK development, and zero-copy hardware memory pipelines.
What it would actually take: The system requires embedded C++ or Rust using the Rockchip RKNN-Toolkit2 C API to interface directly with the 6 TOPS NPU across multiple cores. Crucial hard parts include multi-threaded inference pipelining, converting PyTorch weights into quantized RKNN models without loss of mAP, zero-copy V4L2 camera capture via DRM/DMA-BUF, and custom non-maximum suppression (NMS) on the CPU/NPU. This demands specialized expertise in embedded systems, computer vision, and edge hardware acceleration.
Discussion
9 comments analyzed.
Competitors mentioned: RF-DETR, YOLOv9-12, YOLOv5
Concerns raised: NPUs underutilized due to lack of programming knowledge, YOLOv8 is 3 years old; newer versions have murky licensing or aren't open source, Batching adds queueing latency for real-time live camera streams, RKNN export compatibility issues with newer YOLO versions
Competitors
Other products that read as similar to this one — 400 launches clear the similarity bar, closest 8 shown.
Attention rank: #58 of 401 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 226 days after the earliest competitor.
- jev-drone · github · 2026-09-16 · 85 upvotes · similarity 0.57
- mountain-flood-uav-optimization · github · 2026-09-24 · 102 upvotes · similarity 0.57
- Introduction-to-Unmanned-Aircraft-Systems · github · 2026-09-13 · 9 upvotes · similarity 0.57
- DroneTector: high-resolution radars to detect small drones · yc · 2026-02-19 · 10 upvotes · similarity 0.57
- TIO-Former · github · 2026-09-14 · 16 upvotes · similarity 0.56
- FPGA soft-core of the Saab Viggen's 1963 airborne computer · hn · 2026-03-20 · 22 upvotes · similarity 0.52
- Vernius Systems - Autonomous radar guidance for interceptor drones · yc · 2026-08-24 · 9 upvotes · similarity 0.52
- Batear · hn · 2026-03-21 · 6 upvotes · similarity 0.50
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a hardware & robotics tool for Fintech yet.