Retinal Microsurgery is performed with tiny surgical instru- ments which are observed through a microscope. These instruments can have di erent tip's shape based on their types. The divesity of these types makes tracking and detecting of these instruments very complicated pro- cess. Therefore, having prior information about the di erent tool types allows to select the right model to be used in matching, detecting, and tracking. In this work, a new technique is proposed to detect the tool type online, which can be integrated in any tracking and detection based system. To detect the tool-type, a deep convolutional neural network is learned o -line from di erent tools types, and used on-line to predict the type of the tool present within a small box during the live-surgery. The results demonstrated a high accuracy in the tool-type detection.     «

Retinal Microsurgery is performed with tiny surgical instru- ments which are observed through a microscope. These instruments can have di erent tip's shape based on their types. The divesity of these types makes tracking and detecting of these instruments very complicated pro- cess. Therefore, having prior information about the di erent tool types allows to select the right model to be used in matching, detecting, and tracking. In this work, a new technique is proposed to detect the tool...     »