Unmanned aerial systems (UASs)-based spot spraying application is considered a boon in Precision Agriculture (PA). Because of spot spraying, the amount of herbicide usage has reduced significantly resulting in less water contamination or crop plant injury. In the last demi-decade, Deep Learning (DL) has displayed tremendous potential to accomplish the task of identifying weeds for spot spraying application. Also, most of the ground-based weed management technologies have relied on DL techniques to classify weeds from crop plants. However recently, aerial spraying via UASs is also emerging and is in the nascent stage of development. Therefore, to add to the development of aerial spraying application, we are releasing a multiclass UASs acquired weed image dataset called, SpotWeeds. In the past, a lot of weed dataset has made its way on the public domain, but till date no UASs acquired weed dataset has been published that could be used to facilitate aerial spraying application. Primarily, by releasing this dataset we are contributing towards big data for deep learning to leverage aerial spraying application. This image dataset comprises of 6 different types of weed classes, namely, greenfoxtail (Setaria viridis), horseweed (Conyza canadensis), kochia (Bassia scoparia), common ragweed (Ambrosia artemisifolia), redroot pigweed (Amaranthus retroflexus), and waterhemp (Amaranthus tuberculatus). To create this dataset, we have captured a total of 11,100 aerial and greenhouse RGB images using a small Phantom Pro V2 and a hand-held Canon 90D, respectively. To organize SpotWeeds dataset, individual images of each weed class were clipped and saved inside a test, train, and validation folder. Additionally, to add variations to the clipped dataset, we have performed augmentation technique by rotating, shifting, flipping, zooming, and normalizing each image within single class of weeds. After augmenting, a total of 30,815 images have been generated pertaining to a split of testing (10%), validation (20%), and training (70%) subsets. On top of this, we have included high-resolution videos of our test plots by flying UASs at ~10 ft (3 m) height. These videos can be further used to deploy trained DL models to simulate near real-time weed detection. We invite researchers, industrialists, DL experts, and weed scientists to use this dataset for building a robust DL model that can be used to automate weed detection for aerial spraying applications.