PAANI delivers on-device visual evidence fusion for river robots
PAANI is an on‑device perception‑to‑guidance system designed for mobile river monitoring robots. It fuses a YOLO11n object detector and a MobileNetV3 Small semantic segmenter on an Arduino UNO Q, producing explainable advisories that expose the evidence behind each decision.
Key points
- PAANI fuses YOLO11n and MobileNetV3 Small on Arduino UNO Q, 14.817 MB models
- Median latency 467.8 ms, 95th percentile 580.3 ms at 0.5 Hz cadence
- Detector mAP 0.7388, segmenter mIoU 0.9750
The architecture was trained on 10,000 WaterScenes images for four‑class detection and 1,127 MaSTr1325 images for segmentation, with 198 validation images. The FP32 ONNX models occupy 14.817 MB. In tests the detector achieved a mean average precision of 0.7388 at 0.5 IoU, while the exported model scored 0.7367. The segmenter reached a mean intersection‑over‑union of 0.9750. A five‑minute recording on the UNO Q yielded median pipeline latency of 467.8 ms and a 95th percentile of 580.3 ms at a 0.5 Hz cadence.
The study highlights the feasibility of inspectable, reusable edge‑robotics foundations that separate model accuracy from on‑board execution, and it points out issues such as black‑input misclassification and sampling‑rate mismatches that can affect real‑time guidance.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- AI Safety Concerns Erupt Amid Recursive Self-Improvement Fears · 4 src
- Opinion: OpenAI could copy TypeSafe's Jev classifier and embed it in future models · 1 src
- Aws preview TBC rat‑Brain video model for select customers · 2 src
- ChatGPT-6 Astra decodes 108-year-old WWI German cipher · 2 src
- JS Denain of Epoch AI says Chinese models lag US labs by about six to eight months · 1 src
Comments
via GitHub Discussions