A Complex Problem to Solve?
More and more often, classical image processing techniques are not powerful enough to solve an imaging application for which:
- Lighting conditions are varying in terms of intensity and colorimetry;
- A very large set of images has to be automatically analyzed;
- Image acquisition is varying between images;
New innovative Deep Learning techniques help if your application involves too many images to be processed, a whole collection of objects to be analyzed, and many defects to be detected.
Furthermore, long proven techniques such as Image Filtering, or even Image Understanding typically require huge development efforts. The solution to the application is then very specific and needs a lot of tuning as soon as something is changing in the acquisition setup, such as illumination, magnification, small changes in color, or other small changes out of user control.
ADCIS has the Right Solution
Using Deep Learning, a new and innovative technology based on Convolutional Neural Networks, a user can solve applications, such as the recognition of signs along roads and railways, the automatic detection of defective parts, the detection of abnormal cells in the medical field, the assessment of diabetic retinopathy on color fundus images of the eye, and the detection of defects on electric materials. Before using this technology, the list of applications described above did not have a working solution due to the large variability of defects and images.
Since a few years, ADCIS has been developing applications using Deep Learning techniques. The application usually involves first a training phase that requires an expert to manually identify, detect, and classify objects of interest into categories. This phase is called an annotation process. Objects are then stored in a training database that can be shared between experts and programmers. The computer then behaves like a brain, and is capable of automatically recognizing objects of interest.
So, make the right choice, and work with the ADCIS engineering team. A partnership with one of the only worldwide company that masters Deep Learning will guarantee a working solution fully tested and validated.
Benefits from a Customized Deep Learning Solution
- The right tools to be implemented are jointly defined
- An Image Annotation tool is provided to build the learning Database
- The Database is tuned up for efficient learning
- The appropriate Deep Learning method is selected
- A whole set of tools to be integrated in the final application is delivered
- A follow-up of the application is ensured whenever needed