Image compression method and related device

An image compression device, comprising

• a memory configured to store a first threshold value, a second threshold value, a present neural network model and training times of a compression neural network, a compression training image set of the compression neural network, a tag information of each training image in the compression training image set, a recognition neural network model, and a compression neural network model,
• wherein the present neural network model of the compression neural network is determined as a target model,
• wherein the compression neural network model is a corresponding target model if training of the compression neural network is completed, and
• wherein the recognition neural network model is a corresponding neural network model if training of a recognition neural network is completed;
• and
• a processor configured to:
• acquire an original image with a first resolution, wherein the original image is any training image in the compression training image set and a tag information of the original image is determined as target tag information,
• compress the original image based on the target model to obtain a compressed image with a second resolution, wherein the second resolution is lower than the first resolution,
• recognize the compressed image based on the recognition neural network model to obtain reference tag information, and
• acquire a loss function according to the target tag information and the reference tag information,
• wherein if the loss function is convergent to the first threshold value or the count of training times is more than or equal to the second threshold value, the processor is further configured to:
• acquire a target original image with the first resolution, to determine the target model as the compression neural network model, and
• compress the target original image based on the compression neural network model to obtain a target compressed image with the second resolution.