| CPC G06T 5/70 (2024.01) [G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06T 3/4046 (2013.01); G06T 5/73 (2024.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] | 15 Claims |
|
1. An image recognition method, comprising:
receiving an input image of a first quality;
extracting an input feature of a second quality, different than the first quality, from the input image by inputting the input image directly to a trained encoding model in an image recognizing model; and
generating a recognition result for the input image based on the input feature,
wherein the trained encoding model is a neural network-based model that is trained in advance by training a temporary encoding model using a secondary training image set of the first quality and the second quality and using a trained temporary decoding model of which parameters are fixed and trained in advance based on a primary training image set of the second quality, such that the trained encoding model is configured to extract respective input features of the second quality from respective input images irrespective of whether the respective input images are of the first quality or of the second quality.
|