Take the tour and set up CV-Scope

CV-Scope turns a fixed camera or a recorded video into measurements. You draw lines and zones over the picture. CV-Scope then records every time a person, bicycle or vehicle crosses a line or spends time in a zone.

This guide starts the app, runs the first-run setup with a sample video and shows where everything is. The other guides build on it.

Start CV-Scope

Installed with the Windows installer: open CV-Scope from the Start menu. A small CV-Scope window appears, and the browser opens http://127.0.0.1:8420 when the server is ready (the first start can take a minute). Keep that window open while you work: closing it stops CV-Scope and any run in progress. The installer already downloaded the sample videos if you kept Sample videos ticked.

From the source folder:

1. Open a terminal in the CV-Scope folder.

2. Start the server:

   .venv\Scripts\python.exe -m pathscope.cli serve

   On Linux and macOS the command is .venv/bin/python -m pathscope.cli serve.

3. Open http://127.0.0.1:8420 in the browser.

4. Keep the terminal window open while you work. Closing it stops CV-Scope and any run in progress.

The sample videos live in the samples\videos folder. If that folder is empty, download them once:

.venv\Scripts\python.exe scripts\download_samples.py

Run the first-run setup

A new installation opens the setup wizard by itself. You can open it again later at http://127.0.0.1:8420/setup. A second pass only adds another project and camera.

1. System check. It confirms that the API and the database are ready and lists the detector packages. Click Continue.
2. Hardware. It shows the processor, memory, graphics card and free disk space CV-Scope found. Click Continue.
3. Model recommendation. It lists four tiers for this computer: Fast, Balanced, High accuracy and CPU mode. The recommended one is marked. Click Continue.
4. Detector. The recommended detector is already selected. Click its Download button and wait until it shows Installed. Then click Continue.
5. Camera or video. Keep Video file (upload or local file). Under Or start with a sample clip, click people-detection.mp4 (the corridor sample). If no sample clip is listed, type its full path in Or use a file already on this computer, for example C:\Users\you\CV-Scope\samples\videos\people-detection.mp4, and click Use file. The size and frame rate of the clip appear below. Click Continue.
6. Project. Set Project name to CV-Scope practice and Camera name to Corridor. Click Continue.
7. Scene Builder. Check the summary and click Create and open Scene Builder.

The Scene Builder opens on the first frame of the corridor clip. Guide 02 starts from here.

How the pieces fit together

• Project. One study or one monitored place. It groups cameras and experiments.
• Camera. A video file, a USB or built-in webcam, or a network stream. Each camera belongs to one project and has its own scene.
• Scene. What you draw over the camera picture: counting lines, gates, zones, checkpoints, ignore areas, routes and an optional calibration. Once a run has used a scene, saving it creates a new version, so old results stay reproducible.
• Experiment. What to measure: the camera, the scene version, the objects to track, the detector, the tracker and optional rules. Duplicate an experiment to change one condition.
• Run. One execution of an experiment. A run on a video file stops at the end of the file. A run on a live camera continues until you click Stop.
• Events. What a run records: line crossings, zone entries and exits with their duration, route outcomes and rule events. Each event carries an anonymous track id that is only valid within its run.

Where everything is

The sidebar has three groups. In the app, a bold page name in a guide is a link to that page.

The top bar shows API connected while the server answers, and how many runs are active. API unreachable means the server has stopped. Start it again in the terminal.

Next

Guide 02 draws a counting line over the corridor clip you just added and counts the people who cross it.