Mesopelagic and epipelagic fish species are integral to marine biodiversity and represent a significant target for China's marine fisheries. These fish are known for their relatively low connective tissue in the muscle, which makes them more prone to rapid spoilage during transportation, storage, and processing. This vulnerability is due to the action of both endogenous and exogenous decarboxylase enzymes that convert certain amino acids into lots of biogenic amines, leading to the spoilage of the fish. As a result, these species are often referred to as high biogenic amine fish. Under normal circumstances, living organisms contain trace amounts of endogenous biogenic amines, which are essential components of the body and serve important physiological functions. However, an excess of these amines can lead to a variety of harmful effects on the organism, and in severe cases, may even pose a threat to life. The potential hazards posed by biogenic amines in mesopelagic and epipelagic fish are a significant concern for the sustainable development of the industry. The production and control of histamine in fish have become a key area of focus for the development of fishery resources, deep-sea aquaculture, product storage, transportation, and quality control in China. It is also an issue that must be addressed for the development of deep-sea fishing and aquaculture industries. The freshness of fish is crucial to the quality of aquatic products and directly influences their market value. Biogenic amines such as histamine, cadaverine, and putrescine are often found in high concentrations in fish with poor freshness. Therefore, identifying the freshness of high biogenic amine fish is critical for transportation, storage, processing, and food safety. Exploring the changes in sensory indicators of freshness in high biogenic amine fish and understanding how to quickly, accurately, and objectively evaluate their freshness is of great significance to the healthy and sustainable development of the marine fisheries and aquatic product processing industries. This study focuses on two important high biogenic amine fish species in China's coastal waters, the mackerel (Scomber japonicus) and the anchovy (Engraulis japonicus). These species are chosen due to their commercial importance and their tendency to produce high levels of biogenic amines. The study aimed to evaluate sensory indicators at different freshness stages and to develop a reliable and fast method for assessing freshness based on sensory indicators and machine vision. Sensory indicators at different freshness stages were scored and evaluated using a standardized scoring system. The scores were then used to categorize the fish into different freshness levels. In addition, images of the fish body, gills, and eyes at each freshness stage were collected. The PMG (Progressive Multi-Granularity) and YOLOv5 (You Only Look Once version 5) image recognition models were trained using the collected images and sensory evaluation results. These models were then tested for their ability to accurately identify the freshness of the fish under various storage conditions. The PMG model achieved high accuracy rates for the identification of the freshness of anchovy body under room temperature, refrigerated, and frozen storage conditions, with rates of 96.64%, 99.05%, and 98.78%, respectively. Similarly, the model achieved high accuracy rates for the freshness of mackerel bodies, with rates of 99.84%, 97.22%, and 99.78% under the same storage conditions. The YOLOv5 model also performed well, achieving average precision rates of 98.9%, 99.9%, and 99.7% for the identification of the freshness of anchovy eyes and gills under room temperature, refrigerated, and frozen storage conditions, respectively. For mackerel, the model achieved precision rates of 99.7%, 99.8%, and 99.9% under the same conditions. The DenseNet-121 model can achieve a recognition accuracy rate of 74.5% for the freshness of the further refined mackerel and 83.1% for the freshness of the refined anchovy. After data merging, the comprehensive accuracy rate can reach 86.0%. All models showed high consistency with the sensory evaluation results, demonstrating the potential of machine vision-based image recognition models as a reliable method for rapidly identifying the degree of spoilage in high biogenic amine fish. This study provides technical support for evaluating the patterns of quality changes and safety control in high biogenic amine fish. The results are of significant importance for risk early warning of high biogenic amine fish and for the sustainable development and utilization of high biogenic amine fish resources. The implications of this study are far-reaching. By providing a reliable and objective method for assessing the freshness of high biogenic amine fish, it can help ensure the safety and quality of seafood products reaching consumers. Additionally, it can help fisheries and aquaculture businesses maintain a high standard of quality, which can enhance their reputation and customer loyalty. Furthermore, the use of advanced technology such as deep learning and computer vision can help to modernize the seafood industry, making it more efficient and competitive. By automating the process of freshness assessment, businesses can save time and labor costs, while also reducing the potential for human error. This can lead to a more sustainable and profitable industry in the long term. In conclusion, the study on the freshness of mesopelagic and epipelagic fish species using computer vision-based image recognition models is a significant step forward in ensuring the safety and quality of seafood products. It provides valuable insights into the changes in sensory indicators of freshness and offers a reliable method for assessing the freshness of high biogenic amine fish. The results of this study have important implications for the sustainable development and utilization of high biogenic amine fish resources, as well as for the overall health and safety of the seafood industry.