Newest papers

Lightweight Frame Compensation for Real-Time Polyp Segmentation under Frame Skipping

Weihao Gao · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31228 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Transframe skipping, yang melakukan inferensi setiap frame K dan mengulangi atau interpolasi sisanya, meningkatkan output tetapi memperkenalkan kelemahan gerakan dan kerusakan yang merusak konsistensi ruang-waktu. Untuk menutup kesenjangan ini, kami mengusulkan Lightweight Frame Compensator (LFC), sebuah modul Plug-and-play parameter 77K yang dimasukkan sebelum backbone segmensi frame beku. Beroperasi pada jendela tiga frame, LFC menggabungkan frame berturut-turut melalui konsistensi ruang-temporal ke dalam satu frame yang dikompensasi, dan hanya dilatih pada pseudo-labels yang dihasilkan oleh model Transweight yang ditaburkan dengan cara yang berbeda, tetapi tanpa anotasi ahli. Eksperimen pada tiga frame benchmark video polypfright yang menunjukkan LFC mengembalikan akurasi paling akurat dari

Machine translation (Meta M2M100); the English original follows.

Abstract

Real-time polyp segmentation in computer-assisted colonoscopy is hampered by the difficulty of deploying high-accuracy models under strict latency budgets. Frame skipping, which performs inference every K frames and reuses or interpolates the rest, improves throughput but introduces motion blur and lesion misalignment that degrade temporal consistency. To close this gap, we propose the Lightweight Frame Compensator (LFC), a 77K-parameter plug-and-play module inserted before a frozen segmentation backbone. Operating on a three-frame window, LFC fuses consecutive frames via spatio-temporal attention into a single compensated frame, and is trained only on pseudo-labels generated by a differently seeded model, without any expert annotation. Experiments on three polyp video benchmarks show that LFC recovers most of the accuracy degraded by direct frame skipping and, in several settings, reaches the accuracy of frame-by-frame inference, while retaining a 2.05X--2.55X throughput gain. We further evaluate the temporal stability of three backbones spanning CNN, pyramid Transformer, and global-attention Transformer architectures, revealing a spectrum of sensitivity to frame skipping that arises from how locally or globally each architecture aggregates misaligned input. The compensation gain of LFC follows this spectrum: it is largest for high-capacity Transformer backbones, which achieve the best static accuracy but suffer the most under skip-frame inference. LFC thus offers a lightweight, pluggable, and deployment-friendly route to real-time polyp segmentation.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Artificial intelligence, Computer vision, Segmentation, Frame (networking), Motion compensation

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange