Amr-devman/jev-self-driving-vision-demo
User repository on GitHub
Using Jev box-type questions to one-shot vision
- Forks
- 2
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev Vision for Self-driving
A demo that applies Jev box-type questions to traffic footage from Jakarta one frame in four: Jev finds cars, motorcycles and buses, a local tracker fills the other frames, and the app overlays boxes with latency, cost and detection counts.
- Stars
- 4
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- Amr-devman
- Last verified
- Oct 2026
Works forJev for Extraction
Used inJev for Robotics & Hardware
Highlights
- Jev answers box-type questions on traffic video; the app draws boxes and shows latency, cost and detection count.
- Jev checks one frame in four; a local image tracker follows boxes on the other frames to cut API cost and box jitter.
- Tracked boxes are drawn dashed and Jev's boxes solid, per the README's how-it-works section.
- For each checked frame the app sends the full image plus 30 tiles arranged in grids.
- Python 3.11+, ffmpeg and a TypeSafe API key; the input clip is included in the repo.
Quickstart
bash
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
cp .env.sample .env
.venv/bin/python jev_vision.pyWatch out
README reviewed 2026-10-17; this is a small demo (3 stars) with a one-minute Jakarta traffic clip, the digest excerpt cuts off mid-sentence about the 30 tiles, and no detection accuracy figures are given.
Questions about this entry
Short answers drawn from the linked project or page.
A demo that uses Jev box-type questions on a traffic video from Jakarta. Jev finds cars, motorcycles and buses, the app draws boxes around them, and the overlay shows latency, cost and detection count.
Jev checks one frame in four on the source video; a local image tracker follows the boxes on the frames in between, which the README says reduces API cost and box jitter.
Jev's boxes are drawn solid, and boxes followed by the local tracker are drawn dashed, so the video shows which frames were actually answered by the model.
The full image plus 30 tiles arranged in grids, per the README. The digest excerpt cuts off mid-sentence while describing the tiling, so the exact question format is not fully documented.
Python 3.11 or later, ffmpeg, and a TypeSafe API key. The input video is included in the repo's clip/ folder, and the app writes a timestamped output video to out/.
The README shows .venv/bin/python jev_vision.py --limit 57 to try about two seconds of the clip before processing the whole minute.
The GitHub user Amr-devman. The repository was created and last pushed 2026-10-15, with 3 stars and 2 forks when the sweep ran; no license is listed in the metadata.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Happy JevDay: images, box, longer context, and an opt-in to find jaggedness
X: First among models under 10B on the Decision Index v0.2.1. Liquid AI's d1-3B is built from LFM2.5-VL-3B and…
X: The classic hotdog classifier as a real-time vision decision. Fayaz Ahmed rebuilt the not hotdog app on…
X: Liquid d1 with vision
More Jev resources like this
A MuJoCo city simulation where one, two or four camera-observed drones pursue a target: each role asks Jev role-specific typed questions while deterministic mission navigation, safety guards and motor control run independently at 50 Hz.
Jev Demos & Experiments#community#github#python
Community
YongLD/Multi-UAV-JEVPython
YongLD/Multi-UAV-JEVPythonA MuJoCo Franka arm with an Allegro hand stacking a blue cube on a red one, driven two ways: Claude Opus 5 reasoning through low-level tools, and Jev picking each simulated move while code owns the physics.
Jev Demos & Experiments#community#python#robotics
Community
FazalAAli/jev-robotics-demoPython
FazalAAli/jev-robotics-demoPythonA Python library and CLI that tracks objects in real time by combining TypeSafe Jev Box questions with sparse optical flow and a Kalman filter. Jev supplies absolute positions, flow updates velocity every frame, and a history buffer rewinds the filter to the frame each answer describes.
Jev Repos & SDKs#community#github#python
Community
BenCaunt/jev-box-trackingPython
BenCaunt/jev-box-trackingPythonA Python experiment that lets TypeSafe's Jev play Factorio from screenshots alone: each tick the vision model looks at the screen, decides, and presses keys or moves the mouse, aiming to craft 10 automation science packs from a fresh freeplay game.
Jev Demos & Experiments#community#github#factorio
Community
kgaspard/jev-factorioPython
kgaspard/jev-factorioPythonA Python dispatcher skeleton for TypeSafe System One (Jev) that turns typed signatures into batched questions, plus a Playwright browser harness that measures the dispatcher against ground truth.
Jev Demos & Experiments#community#github#python
Community
monkey1sai/jev-systemone-labPython
monkey1sai/jev-systemone-labPythonTwo Claude Code hooks powered by Jev, TypeSafe's System One model: a PreToolUse risk gate that asks three noul questions per tool call and maps the top probability to allow, ask or deny, plus a tier router that rewrites subagent calls to haiku, sonnet or opus.
Jev Demos & Experiments#community#github#claude-code
Community
spmamidi/jev-claude-code-pocPython
spmamidi/jev-claude-code-pocPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.