Trillo VisionAI supports a complete lifecycle of training and inference of vision AI.
It also includes a tool for labeling the data.
- Upload images or videos for processing using SFTP, Trillo File Manager UI, or APIs.
- Label data through an easy-to-use and customizable UI.
- Use Google Vision AI.
- Ask Trillo's team to develop custom models.
- Store results in the database and publishes as APIs.
- Use Trillo Workbench to view results.
- Custom Image Classification
Supports Single label and Multi-label Classification.
Model can be fine-tuned on user-supplied labeled data set of images.
- Custom Object Detection
Multi-object detection per image that supports detection of a large
number of typical objects in an image.
- Custom Image Segmentation
Multi-object Instance and Semantic Segmentation that support
segmenting a large number of classes in both semantic and instance modes.
- Video Feature Extraction, such as
Hand gestures and movements of a speaker.
Different body poses such as left and right tilt of the body.
Eye contact of the speaker in the video.
- Video Similarity
Calculates a similarity score of two given videos that could be used
to identify if the videos are actually the same videos of different ones,
and also to determine how similar they are.
- Video Action Recognition
Classifies several typical actions happening in videos.
- Video Feature Extraction Solution
This is an advanced solution that incorporates Triollo NLP
and Trillo Vision to extract several features from videos
that could be used for many typical use cases including but not limited to:
Text based searching of long duration video content.
Image and video based searching of long duration video content.
Extracting highlights and creating several smaller,
abridged or shortened videos from long duration videos.
Searching specific objects,
people, utterances, actions and events in videos.