The AI approach leverages a model trained on a dataset, hosted on Roboflow. This approach uses the Roboflow API for inference, providing a more advanced and accurate object detection method.
Notes:
The dataset which was provided was the only source of model training, no other dataset or external data was used
Hosting of the model including the training was done on Roboflow for ease of use and better scalability
As of now with a low confidence threshold we are getting good results but this can be improved with the better and more extensive dataset
Non_AI
The non-AI approach uses OpenCV, a powerful computer vision library, to detect objects in images. This method relies on traditional image processing techniques.
Given the nature of the images the output is around 85-90% accurate due to the limitations of the traditional cv approach
Additional Comments
For both of the approaches I have implemented a dataset downloader utility script which automatically download the dataset provided to us in the google drive and picks the first 2 images from each dataset so that we can get results on images from both the datasets. The client_secrets.json file is used for authentication and hence is added as an additional file
Info
Name
Sankalp Singh
Python Version
3.11
Description
AI
The AI approach leverages a model trained on a dataset, hosted on Roboflow. This approach uses the Roboflow API for inference, providing a more advanced and accurate object detection method.
Notes:
Non_AI
The non-AI approach uses OpenCV, a powerful computer vision library, to detect objects in images. This method relies on traditional image processing techniques.
Given the nature of the images the output is around 85-90% accurate due to the limitations of the traditional cv approach
Additional Comments
For both of the approaches I have implemented a dataset downloader utility script which automatically download the dataset provided to us in the google drive and picks the first 2 images from each dataset so that we can get results on images from both the datasets. The client_secrets.json file is used for authentication and hence is added as an additional file