# -*- coding: utf-8 -*- """Noddy.ipynb Automatically generated by Colaboratory. Original file is located at https://colab.research.google.com/drive/1qBNYLNHT87B8kwlwXQ4mMM-0OLDGxtum """ import pandas as pd import numpy as np from datetime import date, datetime from PIL import Image import requests import transformers as trf import torch import math as Math import tensorflow as tf processor = trf.AutoProcessor.from_pretrained("openai/clip-vit-base-patch32") model = trf.AutoModelForZeroShotImageClassification.from_pretrained("openai/clip-vit-base-patch32") #get bert model tokenizer = trf.AutoTokenizer.from_pretrained("fabriceyhc/bert-base-uncased-imdb") model1 = trf.AutoModelForSequenceClassification.from_pretrained("fabriceyhc/bert-base-uncased-imdb") def get_mood(image): #image = image.convert('RGB') #image = image.resize((224, 224)) image = np.array(image) image = np.expand_dims(image, axis=0) inputs = processor(text=["happy", "sad", "angry", "surprised", "disgusted", "calm", "neutral"], images=image, return_tensors="pt", padding=True) outputs = model(**inputs) logits_per_image = outputs.logits_per_image # this is the image-text similarity score probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities return probs def get_productivity(image): #image = image.convert('RGB') #image = image.resize((224, 224)) image = np.array(image) image = np.expand_dims(image, axis=0) inputs = processor(text=["low productivity", "medium productivity", "high productivity"], images=image, return_tensors="pt", padding=True) outputs = model(**inputs) logits_per_image = outputs.logits_per_image # this is the image-text similarity score probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities return probs #from the prob array, get the index of the highest prob and return the label def get_label_mood(probs): labels = ["happy", "sad", "angry", "surprised", "disgusted", "calm", "neutral"] index = torch.argmax(probs) return labels[index] #from the prob array, get the index of the highest prob and return the label def get_label_productivity(probs): labels = ["low", "medium", "high"] index = torch.argmax(probs) return labels[index] #using bert model, get the sentiment of the text def get_sentiment(mood, productivity): text = ["I am feeling " + str(mood) + " and my productivity is " + str(productivity)] inputs = tokenizer(text, return_tensors="pt") outputs = model1(**inputs) probs = outputs.logits.softmax(dim=1) labels = ["negative", "positive"] index = torch.argmax(probs) return labels[index] # Get user input - use image file from local directory, not streamlit # user_upload = Image.open("sad_sid.jpg") # user_upload.resize((224, 224)) # #use the uploaded image to get mood and productivity # mood = get_mood(user_upload) # mood # moodname = get_label_mood(mood) # moodname # productivity = get_productivity(user_upload) # productivity # pr = get_label_productivity(productivity) # pr # #use mood and productivity to get sentiment score using bert base model of huggingface # sentiment_score = get_sentiment(moodname, pr) # sentiment_score # give playlist based on mood and productivity def get_playlist(mood, productivity): if(mood == "happy" and (productivity == "high"or productivity == "medium" or productivity =="low")): return "https://open.spotify.com/playlist/37i9dQZF1DXdPec7aLTmlC" elif(mood == "sad" and (productivity == "high" or productivity == "medium")): return "https://open.spotify.com/album/7xSEO4RLeiTcVuwUFLb3UG" elif(mood == "sad" and (productivity == "low")): return "https://open.spotify.com/album/0qRhFH2PLpHPPYUGxwFwHO" elif(mood == "calm" and (productivity == "high"or productivity == "medium")): return "https://open.spotify.com/playlist/37i9dQZF1DWZeKCadgRdKQ" elif(mood == "calm" and (productivity == "low")): return "https://open.spotify.com/playlist/37i9dQZF1DWWQRwui0ExPn" elif(mood == "angry" and (productivity == "high"or productivity == "medium"or productivity =="low")): return "https://open.spotify.com/playlist/37i9dQZF1DX4sWSpwq3LiO" elif (mood == "disgusted" and (productivity == "high" or productivity == "medium" or productivity == "low")): return "https://open.spotify.com/playlist/37i9dQZF1DXa2SPUyWl8Y5" elif (mood == "surprised" and (productivity == "high" or productivity == "medium" or productivity == "low")): return "https://open.spotify.com/playlist/37i9dQZF1DWXti3N4Wp5xy" else: return "https://open.spotify.com/playlist/37i9dQZF1DX4SBhb3fqCJd" #make an output function that takes in the image and returns the mood and productivity, and then the playlist def out(image): mood = get_label_mood(get_mood(image)) productivity = get_label_productivity(get_productivity(image)) playlist = get_playlist(mood, productivity) sentiment = get_sentiment(mood, productivity) #output fun message based on mood if(mood == "happy"): m = "Yayy! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n You are doing great! Keep it up! \n Here is a playlist for you: " + ""+str(playlist)+"" elif(mood == "sad"): m = "Aww! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n Don't worry, everything will be alright! \n Here is a playlist to cheer you up: " + ""+str(playlist)+"" elif(mood == "angry"): m = "Calm down Angry Bird! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n Take a deep breath and relax! \n Listn to some music: " + ""+str(playlist)+"" elif(mood == "surprised"): m = "Woah! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n All of us love surprieses. Here is another one for you " + ""+str(playlist)+"" elif(mood == "disgusted"): m = "Eww! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n Don't worry, everything will be alright! \n Here is a playlist to cheer you up: " +""+str(playlist)+"" elif(mood == "calm"): m = "Amazing! Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n You are doing great! Keep it up! \n Here is a playlist for you: " + ""+str(playlist)+"" else: m = "Your current mood is " + str(mood) + " and your productivity is " + str(productivity) + ". Your overall sentiment is " +str(sentiment) +". \n Here is a playlist for you: " + ""+str(playlist)+"" return m # c = out(user_upload) # c #import gradio import gradio as gr #image input image = gr.inputs.Image(shape=(224, 224)) #output output = gr.outputs.HTML("

Mood: {mood}

Productivity: {productivity}

Sentiment: {sentiment}

Playlist: {playlist}

") #interface gr.Interface(fn=out, inputs=image, outputs=output, title="Music Recommender - Noddy", description="Upload an image of yourself and we will recommend a playlist based on your mood and productivity level. \n I hope you have fun and I get grades Professor :) \n -Bhumika", allow_flagging=False, allow_screenshot=False, allow_embedding=False, theme="huggingface").launch()