Object recognition in Earth Perception has gone a significant distance in recent years. Object recognition in land photos, in contrast hand, remains a challenging challenge since the objects are recorded minimum pixels and hence are frequently mistaken for the ground. We give the Siri-whu collections for Object Detection on Land Photographs to advance object recognition technology in landscape images. Siri-whu dataset comprises 2400 photos organized into 12 categories. When compared to current land object identification datasets, this dataset delivers a superior answer. As a consequence, we can categorized objects in a picture with more pixel using RESNET-50. It is a layered architecture with the greater number of layers the detection may be increased with high precision and testing results reveal a considerable performance competitive advantage.