The invention discloses an automatic loss assessment method and system for automobile appearance on the basis of a deep neural network. A mobile phone shoots the long-shot/medium-shot photo of an accident automobile, automobile and automobile assembly component detection is carried out on the long-shot/medium-shot photograph in real time, according to an automobile and automobile assembly component detection result, a user is confirmed to shoot in a proper distance, and the user is prompted to shoot a first local automobile photo; the mobile phone continuously shoots the detail photo of the accident automobile, and a damage area in the detail photo is detected in real time; when the damage area is detected, the user is prompted to shoot a first detail photo; after the photo finishes beingshot, the user is continuously prompted to approach to shoot the area, when the damage area is detected again, the user is prompted to shoot a second detail photo, after the second detail photo finishes being shot, a current damage unit is submitted; and a background server carries out damage identification on each group of damage units submitted from the mobile phone, and the identification results of multiple groups of damage units are combined to obtain the damage details of the accident automobile.