Monitoring the development of shrimp has played an important role as it helps to control the process of feeding to enhance the productivity of shrimp farms. In this study, a simple deep learning based method is proposed to estimate shrimp weight using images. In the proposed method, a deep learning model with weights trained on ImageNet dataset is fully trained on images of a shrimp in a pond to estimate shrimp weight. In compared with the existing methods, this method unifies feature extraction and estimation into a single stage. Experiments on five different databases collected from academic and industrial environment are carried out to confirm the effectiveness of the proposed method. Four different deep convolutional neural networks, including VGG16, ResNet50, DenseNet and MobileNet, are utilized to realize shrimp weight estimation task using image. Experimental results show that the highest accuracy of shrimp weight estimation (coefficient of determination) is 100% for academic environment and 96% for industrial environment.