Custom Models

DeepStack provides a simple API to detect common objects in images. The Object detection API supports 80 objects. Often your use case might involve objects that DeepStack doesn’t natively support, or you might want to fine-tune the object detection for your own kind of images, probably CCTV images or night images if the built-in object detection API doesn’t work perfectly enough for you. For this, you can train a new model on your own images and deploy that to DeepStack.

The video above provides end-to-end guide to doing this.

Here we shall go over the full process of preparing your image dataset, training and deploying with DeepStack.

Step 1: Prepare Your Dataset

Step 2: Train Your Model

Step 3: Deploy Your Model

Contents:

• Preparing Your Dataset
• Step 1: Install LabelIMG
• Step 2: Organize Your Dataset
• Step 3: Run LabelIMG
• Change Annotation to YOLO Format
• Step 4: Annotate Your Dataset
• Annotate Your Test Dataset
• Training Your Custom Model
• Option 1: Training on Google Colab with Free GPUS
• Option 2: Training Locally
  • Step 1 : Install CUDA and CUDNN
  • Step 2 : Install Pytorch
  • Step 3 : Clone DeepStack Trainer
  • Step 4: Train your Model
  • Important Parameters
• Deploying Your Model With DeepStack
• Run DeepStack
  • Starting DeepStack
• Run Inference

• Index

• Module Index

• Search Page