Infrared thermography has been widely used in timely detection of overheating defects in cable accessories. However, the traditional manual diagnosis method is time-consuming and laborious, and it relies too much on expert experience. An autonomous diagnosis method, which is based on the Faster RCNN network and the Mean-Shift algorithm, is proposed in this paper. Firstly, the Faster RCNN network is trained to implement the autonomous identification and positioning of the objects to be diagnosed in images (cable terminations and grounding boxes are included in this paper). Then, the Mean-Shift algorithm is used for image segmentation to rapidly and accurately extract the area of overheating. This is achieved via comparing key regions of the three phase accessories. Next the temperature related characteristic parameters of the overheating region are calculated, and the diagnosis results are obtained in accordance with the relevant cable condition assessment criteria. The proposed method has been applied to test against actual infrared images, and results show that the cable accessories and their overheating regions can be located at different shooting angles and under various background conditions. The research helps reduce the dependence on human efforts and expertise and contributes to improving the practice of condition monitoring.