Our full suite of tools help you create a Computer Vision model unique to your needs and optimized for your application.
Your ideas are unique and a pre-trained model may not get you all the way there. No problem: a custom-trained model will take your application to the next level.
Your ideas are unique, and a pre-trained model may not fit your needs. A custom-trained machine learning model will take your application to the next level.
Why Train Your Own
keyboard_arrow_left
Control
The toolkit leverages transfer learning to modify an existing model instead of creating a model from scratch, which saves time. And since your model is trained on data related to your application, you will see increased performance versus a generic model.
Flexibility
Your ideas are unique, and a pre-trained model may not fit your needs. A custom-trained machine learning model will take your application to the next level.
Performance
We give you control over the entire model training pipeline from raw dataset to edge deployable model. We provide base models pre-trained on the COCO Dataset and our pre-tuned hyper parameters allow non-experts to achieve model training success with little experience.
Control
The toolkit leverages transfer learning to modify an existing model instead of creating a model from scratch, which saves time. And since your model is trained on data related to your application, you will see increased performance versus a generic model.
Flexibility
Your ideas are unique, and a pre-trained model may not fit your needs. A custom-trained machine learning model will take your application to the next level.
Performance
We give you control over the entire model training pipeline from raw dataset to edge deployable model. We provide base models pre-trained on the COCO Dataset and our pre-tuned hyper parameters allow non-experts to achieve model training success with little experience.
Control
The toolkit leverages transfer learning to modify an existing model instead of creating a model from scratch, which saves time. And since your model is trained on data related to your application, you will see increased performance versus a generic model.
keyboard_arrow_right
Model Training in 5 Easy Steps
Step 1: Collect Data
The first thing to do is gather data that contains the object or objects you want to detect. Capture videos or images with your phone, scrape the web, or use our image capture app to generate data.
Step 2: Annotate Data
Once you have compiled data, you need to indicate where the objects of interest are. This is done with an annotation tool. We have incorporated CVAT into the alwaysAI platform so all your tools are in one place.
Step 3: Train the Model
Once your data is annotated you are ready to train a custom MobileNet SSD model. Skip the first two steps by using one of our pre-compiled datasets.
Step 4: Export Model
Upload the newly trained model to your personal model repository.
Step 5: Use the Model
Run your personal model on the edge.
Our Image Capture application helps you with your data collection, so that you can collect videos and images directly from your edge device. See this GitHub repo for more.