TOOLS DEVELOPED FOR IMAGE RECOGNITION

The objectives of our R&D

Reduce Time-to-Market

The tools we developed allow our Data Scientists to concentrate on actual high added value tasks for algorithms conception and implementation. We create leverage on three tasks: ease image labelisation process, offer a complete documentation on image recognition libraries, automate algorithms training.

It is the ratio of time a Data Scientist spend on low added value tasks. Our tools bring it down to 20%.

Tools Developed

2017

Launch of our R&D

Dedicated Researchers

Impacts

30%

Extra time for a Data Scientist

100%

Data used for training purposes

Our tools let me focus on the code which requires the highest technical knowledge and the more human intelligence.

Félix Vogeli, Data Scientist @Sicara

Chani, a dedicated tool to visually control a dataset

Build a reliable dataset

Chani, a dedicated tool to visually control a dataset

Build a reliable dataset

Accurate data is the success-driving factor for image recognition projects. Monitoring the quality of data ensures the team that the algorithm is correctly learning. What Chani offers: - 100% of the data is available for AI algorithm training - Data Scientists can create their datasets x2 faster - Data labeling is automated thanks to Deep Learning algorithms (YOLOv3, MaskRCNN)

Ibad, a library of algorithms built for image recognition

Unpuzzle image recognition algorithms

Ibad, a library of algorithms built for image recognition

Unpuzzle image recognition algorithms

Ibad allows data Scientists and project leaders to understand how algorithms train thanks to a state of the art library of image recognition algorithms. More precisely, Ibad allows to: - Refine algorithms to better solve our clients problems - Cut by three the time dedicated to defining the architecture of code - Maintain the state of the art knowledge of the image recognition ecosystem thanks to tutorials and self-learning sessions