 Hello, this is Miguel Genor-Cabrera from the Skolkowo Institute of Design and Technology, Skoltec. I am glad to present our work, DICSPAL, Tilt and Position Rendering Using Poundware Haptic Display and CNN Based Tactile Pattern Recognition. Telemanipulating the formal objects requires high precision and dexterity from the users, which can be increased by kinesthetic and tactile feedback. This work presents a telemanipulation system for plastic pipettes consisting of a multi-contact happy interface link light to deliver happy feedback at the user's palm and two tactile sensor arrays embedded in the two-finger rotary grid ray. We propose a novel approach based on convolutional neural networks to detect the tilt and position while grasping the thermal objects. The CNN generates a mask based on recognized tilt and position data rendered for their multi-contact tactile stimuli during the telemanipulation. Since the resolution of the happy interface is lower than that of the tactile sensor's array, data preprocessing is required to achieve an effective tactile information exchange to the user. In the first stage, the downsized method resizes and adapts the sensor's data array to the dimensions of the happy display with a unique stimulation point per row. The second stage proposes to use a set of predefined tactile patterns as mask arrays, which use depends on the CNN estimation. We have implemented a classification CNN model with two heads for pipette angle and position recognition. To acquire that set, we have used a pipette holder to set up 12 use cases. The classification program includes four classes, and the position classification program has three classes, representing the distal, middle, and proximal sections on the gripper fingertips. In the test accuracy, the underperception model achieves 95%, and the position prediction model 93.98%. The experiment evaluates how the tactile feedback with base mask data improves the user's perception compared to the direct downsized haptic feedback. During the first experiment, the tilt and position perception was rendered directly from the sensor's data to the user's pan by the haptic display using the downsized method. During the second experiment, the CNN classification performed the data masking. Each of the four angles and three position combination was presented five times blindly in random order. Using only the downsized method, the overall recognition rate was 9.67%, and the overall recognition time was 3.97 seconds. The overall recognition rate using the CNN mask was 82.5%, and the overall recognition time was 3.13 seconds. Based on the experimental results, we can conclude that the using a multi-contact tactile feedback on the user's pan combined with the CNN-based rendering methods can potentially improve the tenor manipulation of the floral objects. The proposed system can be applied as remote co-working labs improving the dexterity of the manipulation and the user's response. For further information, you can consult our paper. Thank you very much for your attention.