Object detection, a fundamental challenge in computer vision, involves locating and classifying objects within images. While significant progress has been made, particularly in detecting medium to large objects, the detection of tiny and small objects remains difficult, with performance degrading substantially. This gap presents a critical issue for applications like robotics, autonomous driving, and intelligent surveillance, where accurate detection across all object sizes is essential. In this paper, we introduce TASOD, a novel dataset specifically designed for Tiny And Small Object Detection. TASOD is the first dataset focused exclusively on the challenges posed by small and tiny objects, addressing a significant gap in existing research. Comprising 12,181 images and 207,745 objects across various categories, TASOD provides a large and diverse resource for advancing object detection methods. We conduct a thorough qualitative analysis of several state-of-the-art deep learning architectures on TASOD, offering insights into their strengths and weaknesses. Our analysis highlights the nuanced challenges these models face when dealing with small object detection, and we discuss how these findings can guide future improvements in this critical area. Through this qualitative lens, we aim to foster deeper research and innovation, bridging the performance gap between medium-to-large and tiny-to-small object detection. Code, data and further information are released at https://github.com/LarsFichtel/TASOD .