- UI that can run inference using the model trained unseen and unlabeled images uploaded at the same time.
- Models Can be deployed on any Cloud [ GCP,AWS,AZURE,DO]
- User Can change 15+ parameters of training
- Fast classification using cloud based architecture
This project is self Hosted project.You can host this project on your own servers. Installation steps will be provided soon.
Hosting project on the cloud using docker build was difficult beacuse we haven't done it before so we have to figure how to deploy model on cloud.
Create a platform for training, labeling and deploying and retraining image classification models. Following are the expected features to be deployed: Dataset : link Train a model(s) that classifies images based on the dataset provided.
Create a UI that can run inference using the model trained above on 128 unseen and unlabeled images uploaded at the same time. Once inference is completed, the UI should then be able to visualize these images and their predictions including the confidence score and provide other metrics as appropriate. The UI should have the functionality to change the labels of images that are wrong, add them to a database and run training again. Optionally, the UI should have an option to change the parameters of training. Parameters could be learning rate, number of filters, filter size etc. The newly trained model should be available for use by the UI to run another round of inference
- Recognition : DJCSI Codeshastra 6.0