Adith Manikonda Project / 2025 / AI & ML
BlindSpot Assistive Perception Prototype
Assistive perception prototype for visual context detection, scene parsing, and object cues on edge devices.
Source ↗The Problem
Assistive vision that depends on a round trip to the cloud fails exactly when connectivity does, and streaming a continuous video feed of where someone lives is a privacy problem before it is a latency problem.
The Approach
Processing runs on a Raspberry Pi 5 carried by the user, so the video never has to leave the device to produce a response. A CNN-LSTM handles motion recognition over the frame sequence, and the result is narrated as audio rather than shown, because the output has to work for someone who cannot see the screen.
Measured
640x480 @ 30fps
Capture
0.3MP USB webcam
Raspberry Pi 5
Edge unit
Battery-portable, no rack
What It Does Not Do
Not fully offline yet: some scene description still calls out to cloud services, so the privacy and availability argument only holds for the detection path, not the whole pipeline.
Built With
- Machine Learning
- Raspberry Pi
- Edge Processing
- Computer Vision
More from Adith Manikonda
- 2026 / AI & MLReference Hallucination DetectorReference checking tool for citation-grounded text review, claim tracing, and hallucination evaluations.
- 2024 / AI & MLThredditThreddit is a Chrome extension that improves productivity with non-intrusive nudges powered by the Gemini API.
- 2025 / AI & MLEcoSyncEcoSync is an AI-based sustainability solution for textile industries evaluating resource consumption.