Rw-Haze: A Real-World Benchmark Dataset To Evaluate Quantitatively Dehazing Algorithms

in recent years, plant identification has become increasingly important in agriculture. However, people have suffered from the problem of low identification accuracy for few years because of large intra-class variance, small inter-class variance and irrelevant features introduced by complex background. This paper proposes an attention-based residual convolutional neural network with group convolution (ARG-CNN) for plant identification. ARG-CNN exploits attention mechanism, group convolution, and mixup to improve the discriminative power of features. To validate the proposed ARG-CNN, we collect a large scale plant dataset: LZU200. Experimental results over LZU200 show that ARG-CNN achieve better accuracy as compared with state-of-the-art methods.