ThiNet: Pruning CNN Filters for a Thinner Net.

This paper aims at accelerating and compressing deep neural networks to deploy CNN models into small devices like mobile phones or embedded gadgets. We focus on filter level pruning, i.e., the whole filter will be discarded if it is less important. An effective and unified framework, ThiNet (stands for "Thin Net"), is proposed in this paper. We formally establish filter pruning as an optimization problem, and reveal that we need to prune filters based on statistics computed from its next layer, not the current layer, which differentiates ThiNet from existing methods. We also propose "gcos" (Group COnvolution with Shuffling), a more accurate group convolution scheme, to further reduce the pruned model size. Experimental results demonstrate the effectiveness of our method, which has advanced the state-of-the-art. Moreover, we show that the original VGG-16 model can be compressed into a very small model (ThiNet-Tiny) with only 2.66 MB model size, but still preserve AlexNet level accuracy. This small model is evaluated on several benchmarks with different vision tasks (e.g., classification, detection, segmentation), and shows excellent generalization ability.

Transgenic Centipedegrass (Eremochloa ophiuroides [Munro] Hack.) Overexpressing S-Adenosylmethionine Decarboxylase (SAMDC) Gene for Improved Cold Tolerance Through Involvement of H2O2 and NO Signaling.

Centipedegrass (Eremochloa ophiuroides [Munro] Hack.) is an important warm-season turfgrass species. Transgenic centipedgrass plants overexpressing S-adenosylmethionine decarboxylase from bermudagrass (CdSAMDC1) that was induced in response to cold were generated in this study. Higher levels of CdSAMDC1 transcript and sperimidine (Spd) and spermin (Spm) concentrations and enhanced freezing and chilling tolerance were observed in transgenic plants as compared with the wild type (WT). Transgenic plants had higher levels of polyamine oxidase (PAO) activity and H2O2 than WT, which were blocked by pretreatment with methylglyoxal bis (guanylhydrazone) or MGBG, inhibitor of SAMDC, indicating that the increased PAO and H2O2 were a result of expression of CdSAMDC1. In addition, transgenic plants had higher levels of nitrate reductase (NR) activity and nitric oxide (NO) concentration. The increased NR activity were blocked by pretreatment with MGBG and ascorbic acid (AsA), scavenger of H2O2, while the increased NO level was blocked by MGBG, AsA, and inhibitors of NR, indicating that the enhanced NR-derived NO was dependent upon H2O2, as a result of expression CdSAMDC1. Elevated superoxide dismutase (SOD) and catalase (CAT) activities were observed in transgenic plants than in WT, which were blocked by pretreatment with MGBG, AsA, inhibitors of NR and scavenger of NO, indicating that the increased activities of SOD and CAT depends on expression of CdSAMDC1, H2O2, and NR-derived NO. Our results suggest that the elevated cold tolerance was associated with PAO catalyzed production of H2O2, which in turn led to NR-derived NO production and induced antioxidant enzyme activities in transgenic plants.

Selective Convolutional Descriptor Aggregation for Fine-Grained Image Retrieval.

Deep convolutional neural network models pre-trained for the ImageNet classification task have been successfully adopted to tasks in other domains, such as texture description and object proposal generation, but these tasks require annotations for images in the new domain. In this paper, we focus on a novel and challenging task in the pure unsupervised setting: fine-grained image retrieval. Even with image labels, fine-grained images are difficult to classify, letting alone the unsupervised retrieval task. We propose the selective convolutional descriptor aggregation (SCDA) method. The SCDA first localizes the main object in fine-grained images, a step that discards the noisy background and keeps useful deep descriptors. The selected descriptors are then aggregated and the dimensionality is reduced into a short feature vector using the best practices we found. The SCDA is unsupervised, using no image label or bounding box annotation. Experiments on six fine-grained data sets confirm the effectiveness of the SCDA for fine-grained image retrieval. Besides, visualization of the SCDA features shows that they correspond to visual attributes (even subtle ones), which might explain SCDA's high-mean average precision in fine-grained retrieval. Moreover, on general image retrieval data sets, the SCDA achieves comparable retrieval results with the state-of-the-art general image retrieval approaches.