Abstract: A battery explosion-proof valve, a battery plug-in box and an electric apparatus are provided. The battery explosion-proof valve includes a fire-fighting medium pipeline, a housing and a piercing assembly. The fire-fighting medium pipeline has a medium outlet and a protective film connected to the medium outlet to block a fire-fighting medium in the fire-fighting medium pipeline. The housing has a cavity, a first opening and a second opening. The first opening and the second opening are communicated with the cavity. The fire-fighting medium pipeline is fitted in the first opening. The cavity is configured to be communicated with an inner cavity of a battery plug-in box through the second opening. The piercing assembly is arranged in the cavity and is movable relative to the fire-fighting medium pipeline to pierce the protective film.

Abstract: The present disclosure discloses an immersion valve assembly and an immersion-type secondary battery device with the same. The immersion valve assembly includes a fire-fighting liquid storage chamber including a fire-fighting liquid outlet facing a battery cell module; and an immersion valve. An opening/closing cooperation relationship is formed between the fire-fighting liquid outlet and the immersion valve, and the immersion valve is configured to move under the action of a gas pressure and open the fire-fighting liquid outlet when thermal runaway occurs to the battery cell module. The immersion-type secondary battery device includes a box cover, a box body, a battery core module and an immersion valve assembly that is arranged inside the box cover.

Abstract: Machine learning object-searching methods and apparatuses are disclosed. The method comprises: selecting a state from a set of states of a target object-searching scene as a first state; obtaining a target optimal object-searching strategy whose initial state is the first state for searching for a target object; performing strategy learning by taking the target optimal object-searching strategy as a learning target to obtain an object-searching strategy by which a robot searches for the target object, and adding the obtained object-searching strategy into an object-searching strategy pool; determining whether the obtained object-searching strategy is consistent with the target optimal object-searching strategy; if yes, determining that the strategy learning in which the first state is taken as the initial state of the object-searching strategy is completed; and if not, returning to the step of selecting a state from a set of states of a target object-searching scene.

Abstract: The embodiments of the present application provide an image segmentation method, an image segmentation apparatus, and a fully convolutional network system. The method includes: acquiring a target image to be processed; acquiring image feature data of the target image; inputting the image feature data into a pre-trained target network far image segmentation to obtain an output; wherein the target network is a fully convolutional network comprising a hybrid context network structure, and the hybrid context network structure is configured to extract a plurality of reference features at a predetermined scale and fuse them into a target feature that matches a scale of a target object in a segmented image; and wherein the target network is trained with sample images containing target objects at different scales; and obtaining an image segmentation result for the target image based on the output.

Abstract: The embodiment of the present application provides a machine learning method and apparatus and an object-searching method and apparatus, which are related to the field of artificial intelligence technology and applied to the robot.

Abstract: The embodiments of the present application provide an image segmentation method, an image segmentation apparatus, and a fully convolutional network system. The method includes: acquiring a target image to be processed; acquiring image feature data of the target image; inputting the image feature data into a pre-trained target network far image segmentation to obtain an output; wherein the target network is a fully convolutional network comprising a hybrid context network structure, and the hybrid context network structure is configured to extract a plurality of reference features at a predetermined scale and fuse them into a target feature that matches a scale of a target object in a segmented image; and wherein the target network is trained with sample images containing target objects at different scales; and obtaining an image segmentation result for the target image based on the output.