Home

BDC

Tools for the BDC Satria Data 2026 competition: classifying waste images into Recyclable, Electronic, or Organic.

Pages

• Competition-Overview — rules, dataset, submission format
• Preprocessing-Pipeline — end-to-end data prep flow
• Image-Augmentation — augment_images.py usage
• Image-Sorting — anomaly detection via EfficientNet / CLIP
• Category-Mapping — ImageNet class → category mapping

Repo layout

preprocess/
  augment_images.py       # rotate/blur/brightness/contrast/saturation augmentation
  build_mapping.py        # generates category_mapping.json from imagenet_classes.txt
  classifier.py           # EfficientNet-based classifier
  classifier_clip.py      # CLIP (open_clip) zero-shot classifier
  sort_images.py          # sorts images using classifier.py, flags anomalies
  sort_images_clip.py     # sorts images using classifier_clip.py, flags anomalies
  imagenet_classes.txt    # 1000 ImageNet class names
category_mapping.json     # class idx -> {class_name, category}
prompts.json              # CLIP text prompts per category
petunjuk-teknis.md        # official competition guidelines (ID)

Labels

Code	Category
0	Recyclable
1	Electronic
2	Organic