computer-vision-for-the-chemistry-lab

This project is based on the vessel instance segmentation, which is implemented by @sagieppel.

python main.py

Resource links

Videos Resources link: https://drive.google.com/drive/folders/1xhWR0Rzq6qwpHGN3Rm18-Rby7yZWiOq-?usp=sharing

Weight files link: https://drive.google.com/drive/folders/1SBjdqPskKwYyABW4xGnT_k6p7i7hskcn?usp=sharing

Put weight files in right location:

"utils/Semantic/logs/1000000_Semantic_withCOCO_AllSets.torch"

"utils/InstanceVessel/logs/Vessel_Coco_610000_Trained_on_All_Sets.torch"

Configuration

default mode: GPU mode # mode can be modified in "utils/Vessel_detect.py"

Function Description

The Computer Vision Assisted Experimental Automation Framework is designed for intelligent analysis of experimental phenomena processes, either offline or online. It supports offline video and online RTSP stream inference, and features functionalities such as transparent glass container detection, liquid separation detection, primary color analysis, and color change detection.

The main.py serves as a demo script. Its workflow includes video stream acquisition, video stream ID definition, creation of video analysis objects, return of analysis results, real-time display, and analysis termination. All sub-algorithms (optional) are based on the main algorithm (transparent glass container mask) to extract the ROI area. If no glass container is detected, the algorithm will retry after a specified waiting time until a glass container is found, after which the sub-algorithms can be triggered.

The main directory includes folders such as utils (which contains a series of functions and script tools), Video_Resources (optional, used for offline or testing), and output (automatically generated during program execution). The main directory contains main.py (demo script) and Exp_v3.py (class file).

Details of the script corresponding functions and parameters can be found in the respective comment sections of the code.

Installation Instruction