PinnedHyunji Cha·Jan 15, 2025Data Transformations in Machine LearningWelcome to my first post in medium. I am an aspiring data science analyst and I would like to use this space for my self-growth to learn…
Hyunji Cha·Aug 22DepthFlow vs. TransFlow: Do We Really Need Real Motion for Video Object Segmentation?Video object segmentation has an annoying data problem.
Hyunji Cha·Aug 20[Paper Review] OASIS: Boundary Refinement for Video Object SegmentationModern video object segmentation methods such as XMem, Cutie, and SAM2 rely heavily on memory. The model remembers what the object looked…
Hyunji Cha·Aug 17Inside Qwen3-VL: A Deep Dive into the Architecture Behind Qwen’s Vision-Language ModelFrom pixels to visual tokens to language reasoning — understanding SigLIP-2, the vision-language merger, DeepStack, Interleaved MRoPE, and…
Hyunji Cha·Jul 4, 2025Understanding IP Addresses, Ports, and Protocols: How the Internet Connects EverythingIf you’ve ever wondered what an IP address really is, what ports mean, or what those acronyms like TCP, UDP, HTTP, and FTP actually do —…A response icon1A response icon1
Hyunji Cha·Jan 18, 2025Ensemble LearningMachine learning has revolutionized how we analyze data, making it possible to automatically find patterns and make predictions. However…