Abstract:Flow-matching video generators produce temporally coherent, high-fidelity outputs yet routinely violate elementary physics because their reconstruction objectives penalize per-frame deviations without distinguishing physically consistent dynamics from impossible ones. Contrastive flow matching offers a principled remedy by pushing apart velocity-field trajectories of differing conditions, but we identify a fundamental obstacle in the text-conditioned video setting: semantic-physics entanglement. Because natural-language prompts couple scene content with physical behavior, naive negative sampling draws conditions whose velocity fields largely overlap with the positive sample's, causing the contrastive gradient to directly oppose the flow-matching objective. We formalize this gradient conflict, deriving a precise alignment condition that reveals when contrastive learning helps versus harms training. Guided by this analysis, we introduce DiReCT (Disentangled Regularization of Contrastive Trajectories), a lightweight post-training framework that decomposes the contrastive signal into two complementary scales: a macro-contrastive term that draws partition-exclusive negatives from semantically distant regions for interference-free global trajectory separation, and a micro-contrastive term that constructs hard negatives sharing full scene semantics with the positive sample but differing along a single, LLM-perturbed axis of physical behavior; spanning kinematics, forces, materials, interactions, and magnitudes. A velocity-space distributional regularizer helps to prevent catastrophic forgetting of pretrained visual quality. When applied to Wan 2.1-1.3B, our method improves the physical commonsense score on VideoPhy by 16.7% and 11.3% compared to the baseline and SFT, respectively, without increasing training time.
Abstract:High-resolution aerial imagery allows fine details in the segmentation of farmlands. However, small objects and features introduce distortions to the delineation of object boundaries, and larger contextual views are needed to mitigate class confusion. In this work, we present an end-to-end trainable network for segmenting farmlands with contour levees from high-resolution aerial imagery. A fusion block is devised that includes multiple voting blocks to achieve image segmentation and classification. We integrate the fusion block with a backbone and produce both semantic predictions and segmentation slices. The segmentation slices are used to perform majority voting on the predictions. The network is trained to assign the most likely class label of a segment to its pixels, learning the concept of farmlands rather than analyzing constitutive pixels separately. We evaluate our method using images from the National Agriculture Imagery Program. Our method achieved an average accuracy of 94.34\%. Compared to the state-of-the-art methods, the proposed method obtains an improvement of 6.96% and 2.63% in the F1 score on average.