With waste generation rising, the need for Smart Waste Management (SWM) solutions is be- coming more important for society to pursue sustainable practices and eliminate inefficiencies in the waste cycle. While there are solutions to many waste-related tasks, the technological surge that has followed the introduction of highly capable Machine Learning (ML) models offers exciting possibilities for what can be achieved when applied to domain-specific tasks. This thesis presents a comparison between a currently available laser sensor for estimating the fill-level of waste bins and ML-assisted Monocular Depth Estimation (MDE) methods. As for the ML methods, both cus- tom machine learning methods and a fine-tuned GPT-4o model are used to determine fill-level from images. The comparison is done by looking at the fill-level estimation performance of the systems, as well as comparing the solutions in a SWM framework. The results show that the monocular camera methods can achieve similar levels of performance in the fill-level estimation task while having several advantages over the dedicated sensor system in the SWM framework. These results enable monocular camera fill-level estimation as a viable option for recycling rooms in the future, which could have large positive effects on users in terms of lower operating costs and more organized recycling rooms. However, there are still challenges to solve with monocular fill-level estimation and more studies are needed to optimize recycling rooms further.