Features
The Fishial.AI portal is a crowdsourcing user image portal that has a built in AI/ML labeling editor that uses "Fishial Recognition™" to auto detect a fish in a image an identify the fish species.
Portal Features
Main features
- 100% cost free for anyone to use and contribute to the effort
- Web based multiple user crowdsourcing image portal
- Support teams of users working on a single team image library togather
- Allows uploading fish images and quick tagging for machine learning
- Image upload with fish objection detection automatically creating a segmentation polygon around the fish
- Build in fish AI detection of fish species already in the AI model
- Allows complex search to be saved, shared and reused
- Built in machine learning tagging editor
- Draw a polygon around a fish in the image
- Automatic object selection & polygon creation tool
- Tag fish species in image and other meta tags
- Water depth of fish in water
- See & set image copyrights per image
- Draw a fish total length (TTL) line on fish species
- Draw polygons around various fish traits in selected object
- Tag fish specie in object
- Fish traits helper image below object
- Fish traits supported
- Mouth form
- Caudal Fin
- Pectoral Fin
- Anal Fin
- Head shape
- Dorsal Fin
- Adipose Fin
- Pelvic Fin
- Eye
- Fish Scales
- Finlets
- Patterns
- Lateral Line
- Other traits supported per polygon
- Species Sex
- Species Age
- Species is Spawning
- Weight
- Length
- Girth
- Multiple fish species can be tagged creating one polygon per fish in photo
- Supports Teams
- One or more Individual can create or join a team
- Work on a common set of images
- Export dataset formats
- Standard Microsoft COCO format
- Fishial COCO extended format that includes fields tagged in image that are not define in a standard COCO model
AI Models
Fish Detection
AI model that can accurately detect the existence of a fish in photo
- Targeted for a both the server an mobile device
Fishial Recognition™
AI model that can accurately perform fish identification with 70-90% accuracy
- Initial model runs within a cloud server and is accessible via an API