The dataset

The dataset contains 70 original images, mostly of defective wafers, and 64 augmented images that contain only good wafers. The three wafer classes are all annotated in LabelMe format, as shown below.

Load the annotated dataset

1. Request access to ReliVision Learning Center and download the silicon wafer image set and its LabelMe annotation files.
2. In ReliUI, import the image set with Import Data.
3. Import the annotation file for each image folder with Import Annotated Dataset, choosing the LabelMe annotation format.
4. Your annotated dataset is now available in the Gallery.

Train the wafer detector

The detector finds every wafer in the image, whatever its condition, and outputs a bounding box.

Prepare the training data

1. Make a copy of the annotated dataset for detector training.
2. Merge all wafer labels into a single class called Wafer.

Train the model

1. In ReliTrainer, select Object Detection as the model type.
2. Select the wafer detector training dataset you prepared.
3. Split the data into train and test sets, automatically or manually, to hold back some data for validation.
4. Set the hyperparameters and start training.

Hyperparameters used in this use case

• Epochs: 40 (default)
• X-Y resolution: 704 (default)
• Learning rate: 0.0001 (default)
• Momentum: 0.937 (default)
• Weight decay: 0.0001 (default)

Train the wafer classifier

The classifier looks at one cropped wafer at a time and labels it Good, Small Defect or Big Defect.

Prepare the training data

1. Extract the rectangle annotations with the ReliUI data curation functions, so each wafer becomes its own image.
2. Keep the three labels on the extracted images — this is your classification dataset.

Train the model

1. In ReliTrainer, select Classification as the model type.
2. Select the wafer classifier training dataset you prepared.
3. Split the data into train and test sets, automatically or manually, to hold back some data for validation.
4. Set the hyperparameters and start training.

Hyperparameters used in this use case

• Epochs: 300
• X-Y resolution: 224 (default)
• Learning rate: 0.0001 (default)
• Momentum: 0.99 (default)
• Weight decay: 0.0005 (default)

Build and run the pipeline

Use the pipeline editor to build your pipeline by dragging and dropping the AI/Basic blocks. You will need to select an input data source (your raw image set) and a detection AI Block, a size filter,  a defect classifier (the classifier you trained) and a position filter  connected in series as depicted below.

1. Open the pipeline editor.
2. Add an input data source and point it at your raw silicon wafer image set.
3. Add the wafer detector AI block and connect it after the input.
4. Add a size filter block and connect it after the detector.
5. Add the wafer classifier AI block and connect it after the size filter.
6. Add a prediction position filter and connect it after the classifier.
7. Add an Output Imageset or End Branch block, depending on the desired output: generate an Imageset file from the executed pipeline or proceed to the audit stage.
8. Run the pipeline with the execute button at the top.

Review the results

Review the results in the ReliUI data annotation interface. The pipeline saves its outputs as a separate set of annotations for each image it runs on, so you can compare predictions against your labels image by image.