Abstract: The present disclosure relates to a coarse annotation based semantic segmentation model training method and device operating by sending an original image into a semantic segmentation model for processing; acquiring a semantic feature map output by a specified convolution layer in the semantic segmentation model; sending the semantic feature map to a first training branch for training to obtain a first cross entropy loss value, wherein the first training branch is a fully supervised branch depending on a ground-truth map; sending the semantic feature map to a second training branch for training to obtain a second cross entropy loss value, wherein the second training branch is an unsupervised branch without a ground-truth map; and determining an overall loss function according to the first cross entropy loss value and the second cross entropy loss value.

Abstract: A heat exchanging device includes: an inner wall and an outer wall, wherein the inner wall is close to the center axis of the heat exchanging device. The inner wall and the outer wall together form a chamber for a cooling medium to flow. The inner wall is provided with at least one protrusion component having an internal cavity. The protruding direction of the protrusion component faces the center axis. The internal cavity of the protrusion component is in communication with the chamber formed by the inner wall and the outer wall. The protruding direction of the protrusion component faces the crystal bar, and the internal cavity of the protrusion component is in communication with the chamber formed by the inner wall and the outer wall, which increases the heat exchanging area, and reduces the horizontal distance between the cooling medium and the crystal bar.