Hello! Today we will tell and show how machine learning and computer vision once again help in solving various problems. This time, our team took part in a case from Finkase LLC as part of the Digital Breakthrough competition of the North Caucasian IT Hub.
We were offered to develop a prototype of an intelligent system for determining the quality of apartment renovation based on computer vision algorithms using Intel tools - OpenVINO ( Open Visual Inference & Neural Network Optimization ).
Case:
When evaluating any real estate object, we are faced with the task of determining the quality of apartment renovation. The quality of finishing is one of the important parameters of pricing, which, unfortunately, is often not indicated in the information about the object. It is required to develop an assessment algorithm that allows to determine the presence of repairs and the quality of finishing from a photograph for subsequent use of the result when assessing the value of objects.
: ( , , ), – . Resnet50. 50 , 12500 . ONNX, ONNX OpenVINO.
ONNX OpenVINO Model Optimizer :
python3 mo.py --input_model <INPUT_MODEL>.onnx
OpenVINO. 93%. , Resnet152 ( , , ).
, Monk. . !
. MIT ADE20K.
2 , 93% (, , , ) (, , , ). . , . . API , , .
, . : , . .
ISUvision ( , , , , 19--1, ) – « ».
Thanks for attention! We advise you to look into our other article and get acquainted with our experience of using Intel tools - OpenVINO not only on hackathons, but also for solving real business problems.