Despite the increasing need to catalogue and describe biodiversity and the ecosystem processes it underpins, these tasks remain inherently challenging. This is particularly true for species that are difficult to observe in their natural environment, such as fossorial and cryptic crabs that inhabit intertidal sediments. Traditional sampling techniques for intertidal crabs are often invasive, labour intensive and/or inconsistent. These factors can limit the amount and type of data that can be collected which in turn hinders our ability to obtain reliable ecological estimates and compare findings between studies. Computer vision and machine learning algorithms present an opportunity to innovate and improve sampling approaches. Moreover, cheaper and tougher recording devices and the diversity of open source software further boost the possibilities of achieving rigorous image-based sampling, which can broaden the range of questions that can be addressed from the data collected. Despite its significant potential, the software and algorithms associated with image-based sampling may be daunting to researchers without expertise in computer vision. Therefore, there is a need to develop protocols and data processing workflows to showcase the value of embracing new technologies. This paper presents a non-invasive computer vision and learning protocol for sampling fossorial and cryptic crabs in their natural environment. The image-based protocol is underpinned by fit-for-purpose and off-the-shelf software. We demonstrate this approach using fiddler crab and sediment recordings to study and quantify crab abundance, motion patterns, behaviour, intraspecific interactions, and estimate bioturbation rates. We discuss current limitations in this protocol and identify opportunities for improvement and additional data stream options that can be obtained from this approach. We conclude that this protocol can overcome some of the limitations associated with traditional techniques for sampling intertidal crabs, and could be applied to other taxa or ecosystems that present similar challenges. We believe this sampling and analytical framework represents an important step forward in understanding the ecology of species and their functional role within ecosystems.