Eployed the model to a new dataset for testing. They identified that the generalization potential from the model is just not high. This also shows the challenge on the underwater DMPO manufacturer environment to a specific extent. Knausgard et al. [235] combined the two tasks of fish detection and fish classification and proposed a phased in-depth understanding strategy for the detection and classification of tropical fish: in the very first stage, Yolo three was applied to detect fish bodies, and within the second stage, CNN-SENet was utilized to classify the detection outcomes in the prior stage. Our function is related to this, but we use phased rotating box object detection and pose estimation, and the output is the integration with the benefits of your two stages. These functions have not organically combined the mature object detection model and human pose estimation model within the current deep finding out method and applied them to fisheries. Our operate is committed to filling this gap. Even so, the construction of an intelligent aquaculture method has been challenged and hindered to some extent. Firstly, the complicated underwater natural environment for example the growth of algae and uneven distribution of light has caused some obstacles towards the collection of visual information of aquatic animals [26]. Secondly, attitude estimation ordinarily takes humans and autos with restricted attitude modifications because the target objects [27,28]; Even though aquatic animals have no limb movement, their movement inside the water is much more open, can flip freely, and is just not restricted by angle. The role of frequent data annotation becomes really limited. To meet the above challenges, we use multi-object detection and animal pose estimation, real-time monitoring, early warning, and recording productive details to decrease the loss. Within this regard, the aquatic animal we mostly study is definitely the golden Eggmanone Purity crucian carp. Based on its inherent benefits, this species plays a far more distinctive part:Fishes 2021, six,3 of(1)(2)(three)(four)The physiological structure of golden crucian carp is comparatively basic, there are actually no complex human-like joints in addition to a higher degree of freedom limbs, as well as the purposeful grass goldfish has high attitude recognition. Including spawning, eating, skin infection, etc. Even though the body look similarity of golden crucian carp is high, the dataset according to artificial annotation was screened and analyzed, as well as the source is trustworthy, which is explained in detail in Sections two.1 and 2.two. The ecological fish tank using a higher reduction degree has a higher simulation on the aquaculture environment. In contrast, it truly is additional in line using the specifications from the aquaculture industry chain, has no redundant interference, and may be freely captured from all perspectives. Golden crucian carp can comprehend free movement in three-dimensional space inside the aquatic environment. In line with Figure 1, the turnover range of golden crucian carp is among [0 180 ]. Commonly, the deformation degree is huge. As shown in Figure 2, 80 from the angle adjustments are above 40 degrees. For that reason, the traditional object detection pre-selection box is abandoned, along with the rotating box is made use of for flexible box choice. This is the innovation on the dataset in our analysis course of action.Figure 1. Analysis of crucian carp dataset. This figure is really a heat map on the x, y, and width, height from the crucian carp image. The darker the colour, the stronger the concentration, and the denser the distribution of crucian carp.Figure two. Evaluation of crucian carp dataset. The angle distribution histogram.