The Bounding Box is the simplest type of annotation, probably the most common. The complexity of Bounding Boxes labeling tasks is often underestimated - a lack of precision can make learning more difficult or time consuming. Innovatiana's Data Labelers are trained in the best annotation techniques - our approach, which includes mandatory training and quality review, ensures the highest level of quality.

An annotation format close to the Bounding Box... but in three dimensions! Particularly useful for your AI products if you work in the automotive industry (but not only!).

To ease the training of your models, you can choose to annotate objects with polygons, delineating the objects very precisely to eliminate noise. This will prevent you from annotating irrelevant elements that could make your model confusing. This of course takes a little more time... but good news, our Data Labelers have been trained in the best tools to label polygons in a reasonable time.

What more can be said? These are dots - on pictures. What for? Often to train detection models or facial recognition. To detect emotions, expressions, ... It's a precision work that requires rigor and resilience ... qualities that characterize our Data Labelers!

Lines, to delimit sections on an image and train your "Computer Vision" model to recognize and delimit roads, streets, sidewalks, ... Because even if your object detection model for your autonomous car is very good, nobody wants his car to confuse a sidewalk with a tree. That's what Lines & Polylines are for.

We are regularly asked to categorize video sequences to train models or algorithms. Our Data Labelers are used to doing this with the most powerful tools on the market for this type of use case, such as V7. To allow you to keep only useful sequences, eliminate noise and structure your video data.

Do you have thousands of unused images? We can classify them and associate semantic attributes to these classes to allow you to filter / search your images in a fluid way... or train a model to do it for you! Do you have a simple case that requires you to categorize 1'000 images in 3 different categories? A complex case where you need to categorize 40'000 images into 40 classes and 50 attributes? Contact us, we've already done it!

Segments, to generate masks on a multitude of images, by managing occlusion or overlays. A work that requires patience, rigor and the use of powerful tools. If you are not equipped, we recommend you to use CVAT... and to call our Data Labelers who master this tool very well!

LiDAR (3D Point Cloud) annotation is a complex task, which requires Data Labelers to be trained and to use efficient Data Labeling tools. For this type of use case, we set up taskforces of experienced Data Labelers, led by a Data Labeling Manager who is an expert in the field.