Behavior is an important parameter that reflects the health and mental state of the mouse. A variety of experimental tools/procedures including running wheel test and open field test, have been developed to evaluate mouse behaviors. We have been developing the methods to assess mouse behaviors under non-invasive and unrestrained condition, using image analysis, and artificial intelligence. We first aimed to establish mouse tracking system to assess its spontaneous movement. Mice were housed in a standard cage in a 12 hour:12 hour light / dark environment. The infrared lamp was used during the dark period. The mouse behavior was continuously recorded from above using a video camera. The geometric center of the mouse in each frame of the captured video was calculated, and its movement was expressed as the amount of exercise. We confirmed that the mouse movement in the dark period was larger than that in the light period as is reported. Administration of caffeine increased the mouse movement, while administration of chlorpromazine, a sedative disappeared it. We next aimed to establish a novel method to detect scratching using deep neural network. Mouse scratching behavior was recorded by video camera. Images showing differences between two consecutive frames in each video were generated, and each image was labelled as showing scratching behavior or not. A convolutional recurrent neural network (CRNN) was constructed and trained using labelled images, and then its performance was evaluated. The predicted number and durations of scratching events were correlated with those of the human observation. Thus, we succeeded in establishing methods that can analyze the spontaneous movement or scratching of mice using a versatile video camera in a state close to a normal breeding environment.