Music Signal Processing has significantly evolved in the past decades. One of the major areas of interest in this field has been automatic music transcription. It is a challenging task by itself that aggravates even more when the input audio is polyphonic (multiple instruments and timbres are played simultaneously). This requires the identification of musical instruments in the piece at the outset. The field of music signal processing that deals with this aspect is known as automatic music instrument identification. This field also has the potential of categorizing and recommending music based on instruments. Disparate datasets have been proposed to date for this task but none of them have interclass background similarity to the best of our knowledge. Further, the lead melody being played also varies from class to class in most cases. These aspects can introduce a possible bias for the machine learning models that can get manipulated unfairly by the additional variances in the classes other than the lead instrument itself. This sets the stage for a dataset where the classes are different only in terms of the tone of the lead instrument alone. In this paper, we introduce the first musical instrument dataset of 10 musical instruments with Electronic Dance Music melodies (EDM10) having identical background music (BGM) across instruments. This dataset is the first of its kind wherein synthetic tones have been used that have taken over the Globe. We introduce out-of-mood testing using exotic scales for musical instrument identification. The dataset is composed of 35800 polyphonic clips of 3 s each and a baseline result of 89.73% was obtained using a deep learning-based approach. The dataset is freely available for research purposes. https://forms.gle/yV5e36TK1jHMKjpC6 .