Unverified50% confidenceFactExact time
The convolution operation uses a small weight matrix (convolution kernel) to slide over the image, computing a weighted sum over a small region each time, and the output is called a feature map; different convolution kernels can detect different patterns.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe principle of tiled processing is to split the feature map into sub-tiles along the spatial dimension for independent computation and then merge them; peak VRAM usage decreases linearly as the number of tiles increases, but overlapping sampling is required to avoid seams and the total computation increases slightly.75% similarUnverified单应性矩阵是3×3的齐次变换矩阵，具有8个自由度，描述同一平面在两个不同相机视角下的投影关系，通过估计四对点对应关系即可唯一确定72% similarUnverifiedSpectral clustering maps high-dimensional data to a low-dimensional spectral space by computing the eigenvectors of the graph Laplacian matrix, and then completes clustering with k-means.72% similarUnverifiedFor crowded/overlapping chick scenarios, instance segmentation can be used to separate overlapping targets via pixel-level masks, or when a detection bounding box area exceeds the average area of a single chick, the count can be estimated by dividing the area proportionally.71% similarUnverifiedBounding boxes use only rectangles to locate targets, with low annotation cost and fast speed, suitable for scenes with little occlusion between targets; segmentation masks outline target contours pixel by pixel, enabling precise differentiation of overlapping targets.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1018398API
curl https://kongchang.com/api/v1/knowledge/claims/1018398MCP
get_claim(id=1018398)