Recommended Conditions for Target Images
Image Tracker recognizes and tracks images as feature points of the target image; therefore, you can experience reliable and steadfast recognition and tracking performance without any jittering by using images with clearly distinguishable feature points.
When you upload a target to Target Manager, the server will proceed with the learning session, and the learning machine analyzes the image to extract the feature points, and classifies the augmentation level of the target according to the number and distribution of the extracted feature points.
The augmentation level of an image is indicated by ★ to ★★★★★ depending on the recognizability and traceability. An image rated as ★ may not be recognizable, namely, it means that as the number of stars increases, the number of feature points extracted is evenly distributed. Therefore, it is recommended to use images with high augmentation level. Images that are difficult to find feature points may not be learned by the server. In this case, you will need to upload a different target image.
Images with ambiguous boundaries between lines and faces are hard to recognize because it is difficult to extract the feature points from them.
The above image was rated ★.
If there is a feature point in only a part of the image, it will not be recognized well.
The above image was rated ★.
If the image consists of patterns, it will not be recognized as a repetition of similar feature points.
The above image was not learned by the server.
Since the image contains a variety of faces and lines, it can be clearly distinguished. You can expect a better recognition performance if the feature points are spread evenly.
The above image was rated ★★★★★.
The learning data of the target image not learned by the server due to the few feature points are not downloaded.
It is recommended that you use a high-quality image as a target, since it is not well recognized when using a low-grade image as the target.