User repository on GitHub
Realtime object tracking with TypeSafe Jev Box questions, optical flow, and a Kalman filter
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
Jev Box Tracking
A 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.
- Stars
- 5
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 2 min
- Published by
- Community
- Author
- BenCaunt
- Last verified
- Oct 2026
Highlights
- Jev gives an accurate box but slowly; Lucas-Kanade optical flow with a RANSAC similarity fit fills the gaps between answers.
- A Kalman filter fuses both sources: flow updates velocity every frame while Jev corrects absolute position.
- When an answer for frame k arrives, the filter rewinds to k, applies the Jev box, then re-runs over frames k+1 to now.
- Install from git and set TYPESAFE_API_KEY; the webcam demo runs jev-box-track "iPhone" to track a named object.
- TrackResult carries present, confidence, latency, replay_ms, tracked points, raw Jev answers and any request error.
Quickstart
bash
pip install "git+https://github.com/BenCaunt/jev-box-tracking.git"
export TYPESAFE_API_KEY=...
jev-box-track "iPhone"Watch out
README reviewed 2026-10-17; it is short and shows no accuracy numbers or failure modes beyond the rewind design, and the tracker depends on TypeSafe's Box questions and a TYPESAFE_API_KEY, so latency and correctness depend on the API round trip.
Questions about this entry
Short answers drawn from the linked project or page.
It is a Python library and CLI for real-time object tracking that uses TypeSafe Jev Box questions together with sparse optical flow and a Kalman filter, so Jev supplies the accurate absolute box while fast frame-to-frame motion comes from image features.
Jev answers Box questions about frames asynchronously, one answer per API round trip. Each answer corrects the tracked object's absolute position in the filter, with the raw Jev answers available on every TrackResult.
Jev gives an accurate box but slowly, since one answer arrives per API round trip. Sparse Lucas-Kanade flow with a RANSAC similarity fit measures motion between answers, and the Kalman filter fuses both so velocity updates every frame.
Install with pip install "git+https://github.com/BenCaunt/jev-box-tracking.git", set TYPESAFE_API_KEY, then run jev-box-track "iPhone" and press q to quit. An optional [dotenv] extra loads the key from a .env file.
Jev answers describe frames from the past, so every frame and the filter state after it are kept in a history buffer. When an answer for frame k arrives, the filter rewinds to frame k, corrects with the Jev box, then re-runs over frames k+1 to now.
TrackResult carries present, confidence, latency, replay_ms, the tracked feature points, the raw Jev answers and any request error, plus a box in image fractions and a pixel_box helper. The update call never blocks on the network.
It is published by the GitHub user BenCaunt under the Apache-2.0 license; the repository was created 2026-10-16 and pushed the same day.
More Jev resources like this
A Python library shaped like the re module that finds, replaces, splits and extracts text by meaning: loose finders propose spans and Jev judges each one with a probability. Every match stays a verbatim slice of the input with its offsets, so it can be wrong but never invented.
Jev Repos & SDKs#community#github#python
Community
nothans/ex-regexPython
nothans/ex-regexPythonA Jev-powered multimodal mobile automation SDK: a standalone Python tap SDK for Android, iOS and HarmonyOS where you hand over a tap target and the SDK locates and executes it on the current screen, with millisecond-level local processing.
Jev Repos & SDKs#community#github#mobile
Community
dongxinsuperman/jev-phonePython
dongxinsuperman/jev-phonePythonTurn markdown into a knowledge graph with an ontology you write and a cutoff you own. A local decision model answers closed yes-or-no questions one at a time, and anything it is unsure about waits for a person instead of entering the graph.
Jev Repos & SDKs#community#github#rust
Community
raphaelmansuy/edgextractTypeScript
raphaelmansuy/edgextractTypeScriptOpen infrastructure from the SimpleJev project for training, evaluating and deploying System 1 decision models across language and multimodal backbones: one API takes a state, question and candidate options and returns a choice with probabilities.
Jev Repos & SDKs#community#github#python
Community
SimpleJev/JevAnyPython
SimpleJev/JevAnyPythonA 24/7 market-making system built around Jev: deterministic code computes the market state, Jev answers six atomic judgments with calibrated probabilities, and the policy engine applies thresholds while hard risk vetoes stay in code.
Jev Repos & SDKs#community#github#python
Community
buberlo/jev-traderPython
buberlo/jev-traderPythonA Python library that turns Jev-style decision models into ordinary function calls: check returns a bool, classify picks one option, label picks every option that applies, and score rates on a scale, with probabilities behind every answer.
Jev Repos & SDKs#community#github#python
Community
jlowin/vibecheckPython
jlowin/vibecheckPythonFrom 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.