Abstract: Techniques are generally described for fusing sensor data of different modalities using a transformer. In various examples, first sensor data may be received from a first sensor and second sensor data may be received from a second sensor. A first feature representation of the first sensor data may be generated using a first machine learning model and a second feature representation of the second sensor data may be generated using a second machine learning model. In some examples, a modified first feature representation of the first sensor data may be generated based at least in part on a self-attention mechanism of a transformer encoder. The modified first feature representation may be generated based at least in part on the first feature representation and the second feature representation. A computer vision task may be performed using the modified first feature representation.

Abstract: A set of enhancements to further improve the performance of deep learning artificial intelligence algorithms trained to detect and localize colon polyps. The enhancements spanning training data mining efficiencies and automation, training data augmentation, early detection of polyps enable a more performant colon polyp detection solution for use on colonoscopy procedure recordings or live procedures in endoscopy centers.

Abstract: A set of enhancements to further improve the performance of deep learning artificial intelligence algorithms trained to detect and localize colon polyps. The enhancements spanning training data mining efficiencies and automation, training data augmentation, early detection of polyps enable a more performant colon polyp detection solution for use on colonoscopy procedure recordings or live procedures in endoscopy centers.

Abstract: A system and method for identifying key frames of a presentation video that include stationary informational content. A sequence of frames is obtained from a presentation video and differences of pixel values between consecutive frames of the sequence of frames are computed. Sets of consecutive frames that are stationary are identified, wherein consecutive frames that are stationary have a proportion of changed pixel values below a first predetermined threshold, and wherein pixel values are deemed to be changed when the difference between the pixel values for corresponding pixels in consecutive frames exceeds a second predetermined threshold. Next, a set of key frames that include stationary informational content is retained. The set of key frames that include stationary informational content is then displayed for user interaction.

Abstract: A system and method for identifying key frames of a presentation video that include stationary informational content. A sequence of frames is obtained from a presentation video and differences of pixel values between consecutive frames of the sequence of frames are computed. Sets of consecutive frames that are stationary are identified, wherein consecutive frames that are stationary have a proportion of changed pixel values below a first predetermined threshold, and wherein pixel values are deemed to be changed when the difference between the pixel values for corresponding pixels in consecutive frames exceeds a second predetermined threshold. Next, a set of key frames that include stationary informational content is retained. The set of key frames that include stationary informational content is then displayed for user interaction.