Interactive videos can help to draw audiences' attention to certain content in a video, resulting in a more appealing or intriguing type of video. Object detection is a fundamental approach that can be used to automatically create interactive videos. In this thesis, we propose a new method for object detection based on combining object recognition with tracking in a neural network. Specifically, we use GoogLeNet [51] as a feature extractor, and then apply a long short-term memory (LSTM) network to further adjust the feature vectors extracted by GoogLeNet according to the context of the feature vectors extracted from the previous frame. We feed the output of the LSTM to a classifier and regressor as in the Overfeat network [46], to obtain predicted confidences and predicted bounding boxes. We pre-train the feature extractor on images of the ImageNet [29] and Pascal Visual Object Classes (VOC) [14] datasets. We then evaluate our network on the visual tracker benchmark OTB100 [5], which is a dataset composed of videos with objects from a variety of classes. We use this dataset for training and testing our network. We compare our results to results obtained without tracking, where GoogLeNet [51] is only combined with the Overfeat network [46] for classification and regression. Our model shows a better performance at predicting objects in frames where occlusion and background clutter appear, and results in more consistent object bounding boxes across frames.