One of the most important tasks in growing and trading shrimp is postlarvae and juvenile shrimp counting. In Vietnam, most small-scale shrimp farms/companies today rely on manual and volumetric counting to estimate the number of postlarvae and juveniles for sale to customers. This traditional counting method is time-consuming, labor-intensive, and prone to error. Inaccurate shrimp counting not only causes economic loss to sellers and buyers but also affects production efficiency. This paper presents the design and implementation of a computer vision-based device for automatic counting of juvenile shrimp.The designed shrimp counting device were portable and low cost by taking advantages of the open-source computer vision library OpenCV and the power of now-ubiquitous smartphones. The experimental results demonstrate that the proposed counting approach can provide an average accuracy of over 96% compared to the true values. The average processing time for one counting is a few seconds. This could provide farmers with a shrimp counting approach that offers acceptable accuracy, requires less time and labor cost compared to the traditional manual counting methods.