To perform object detection, we conducted experiments to learn how to create bounding boxes and also perform dataset creation.
- ObjectDetectionBoundingBox : https://github.com/KrisnaPinasthika/ObjectDetectionBoundingBox
- https://www.kaggle.com/datasets/arkadiyhacks/drinking-waste-classification
- https://www.kaggle.com/datasets/deadskull7/cola-bottle-identification
- https://www.kaggle.com/datasets/moezabid/bottles-and-cans
- Creating datasets manually by taking photos through a smartphone camera
- glass bottle
- plastic bottles
- cans
- cardboard
- rubber
- paper
- plastic
- straws
- Looking for references and creating datasets manually by taking photos through a smartphone camera
- Provide a bounding box and label for each data
- Perform analysis for each image that has been given a bounding box
- Divide the data in a ratio of 80 (train) and 10 (test)
- Create TFRecord for train and test data
- Experiment for freezing layer on feature extractor section of MobileNet v2 architecture
- Augmenting image data and Tuning hyperparameter
- Conduct machine learning model training experiments with 50,000 to 75,000 train steps
- Evaluate the machine learning model
- Testing real data using real photos and using a webcam
- Save the model by changing the input to base64
- Deploy the model to Vertex AI
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent