Long-tail distribution is widely spread in real-world applications. Due to the extremely small ratio of instances, tail categories often show inferior accuracy. In this paper, we find such performance bottleneck is mainly caused by the imbalanced gradients, which can be categorized into two parts: (1) positive part, deriving from the samples of the same category, and (2) negative part, contributed by other categories. Based on comprehensive experiments, it is also observed that the gradient ratio of accumulated positives to negatives is a good indicator to measure how balanced a category is trained. Inspired by this, we come up with a gradient-driven training mechanism to tackle the long-tail problem: re-balancing the positive/negative gradients dynamically according to current accumulative gradients, with a unified goal of achieving balance gradient ratios. Taking advantage of the simple and flexible gradient mechanism, we introduce a new family of gradient-driven loss functions, namely equalization losses. We conduct extensive experiments on a wide spectrum of visual tasks, including two-stage/single-stage long-tailed object detection (LVIS), long-tailed image classification (ImageNet-LT, Places-LT, iNaturalist), and long-tailed semantic segmentation (ADE20K). Our method consistently outperforms the baseline models, demonstrating the effectiveness and generalization ability of the proposed equalization losses.Codes will be released at https://github.com/ModelTC/United-Perception.

No abstract

Increasing crop yield per unit of area can be achieved by increasing planting density. However, high-density planting could trigger shade avoidance responses, which cause exaggerated growth and increased susceptibility to various diseases. Previous studies have shown that the rapid elongation of plants under shade (i.e., reduced red to far-red ratios) is regulated by phytochromes and various phytohormones. However, the detailed molecular mechanisms governing the interaction among these signaling pathways are not well understood. Here, we report that loss-of-function mutants of FAR-RED ELONGATED HYPOCOTYL3 (FHY3) and FAR-RED-IMPAIRED RESPONSE1 (FAR1), which encode two homologous transcription factors essential for phytochrome signaling, exhibit an exaggerated shade avoidance phenotype. We show that FHY3 and FAR1 repress plant growth through directly activating the expression of two atypical basic helix-loop-helix transcriptional cofactors, PHYTOCHROME RAPIDLY REGULATED1 (PAR1) and PAR2, and that this process is antagonized by a group of JASMONATE ZIM-DOMAIN proteins, key repressors of the jasmonic acid (JA) signaling pathway, through physical interactions. Furthermore, we show that FHY3 interacts with MYC2, a key transcriptional regulator of JA responses, coordinately regulating JA-responsive defense gene expression. Our results unveil a previously unrecognized mechanism whereby plants balance their growth and defense responses through convergence of the phytochrome signaling pathway and JA signaling pathway under shade conditions.

In response to competition for light from their neighbors, shade-intolerant plants flower precociously to ensure reproductive success and survival. However, the molecular mechanisms underlying this key developmental switch are not well understood. Here, we show that a pair of Arabidopsis transcription factors essential for phytochrome A signaling, FAR-RED ELONGATED HYPOCOTYL3 (FHY3) and FAR-RED IMPAIRED RESPONSE1 (FAR1), regulate flowering time by integrating environmental light signals with the miR156-SPL module-mediated aging pathway. We found that FHY3 and FAR1 directly interact with three flowering-promoting SQUAMOSA-PROMOTER BINDING PROTEIN-LIKE (SPL) transcription factors, SPL3, SPL4, and SPL5, and inhibit their binding to the promoters of several key flowering regulatory genes, including FRUITFUL (FUL), LEAFY (LFY), APETALA1 (AP1), and MIR172C, thus downregulating their transcript levels and delaying flowering. Under simulated shade conditions, levels of SPL3/4/5 proteins increase, whereas levels of FHY3 and FAR1 proteins decline, thus releasing SPL3/4/5 from FHY3/FAR1 inhibition to allow activation of FUL, LFY, AP1, and MIR172C and, consequently, early flowering. Taken together, these results unravel a novel mechanism whereby plants regulate flowering time by integrating environmental cues (such as light conditions) and an internal developmental program (the miR156-SPL module-mediated aging pathway).

No abstract