New Cookery Recipes

new cookery recipes here you will find everything you are interested in

Inverse Cooking: Recipe Generation from Food Images – Part 3

/
/
/
32 Views

[youtube https://www.youtube.com/watch?v=fx7VgUcCg2o&w=560&h=315]

Hi Friends The paper which we are going to discuss is "Inverse Cooking: Recipe Generation from Food Images" Link to the paper and code can be found at the description section

We have divided the whole paper into 5 segments Those are Abstract, Introduction, Approaches, Experiments and Conclusion We have already covered Abstract, Introduction and Approaches sections in last parts The link is provided in the description Lets's start with Experiments

Experiments This section is devoted to the dataset and the description of implementation details, followed by an thorough  analysis of the proposed attention strategies for the cooking instruction transformer Further, the section also quantitatively compare the proposed ingredient prediction models to previously introduced baselines Finally, a comparison of our inverse cooking system with retrieval-based models as well as a comprehensive user study is provided Dataset Train and evaluate our models on the Recipe 1M dataset composed of around 1M recipes scraped from cooking websites

The dataset contains ~720 thousand training, ~155 thousand validation and ~154 thousand test recipes, containing a title, a list of ingredients, a list of cooking instructions and (optionally) an image Removed recipes with less than 2 ingredients or 2 instructions, resulting in ~252 thousand training, ~54 thousand validation and ~54 thousand test samples Implementation Details Resized images to 256 pixels in their shortest side and take random crops of 224 × 224 for training and central 224 × 224 pixels for evaluation Kept a maximum of 20 ingredients per recipe and truncate instructions to maximum of 150 words The models are trained with Adam optimizer until early-stopping criteria is met

All models are implemented with PyTorch Recipe Generation The system highlights the benefits of using both image and ingredients when generating recipes Noticed that generated instructions have an average of 921 sentences containing 9 words each, whereas real, ground truth instructions have an average of 908 sentences of length 12

79 Ingredient Prediction Considered models from the multi label classification literature as baselines,   and tune them for our purposes ie ingredients should be treated as lists or sets Finally proposed set prediction models, which couple the set prediction with a cardinality prediction to determine which elements to include in the set Generation vs Retrieval In this section, the paper compares proposed recipe generation system with retrieval baselines, which is used to search recipes in the entire test set for fair comparison

Results show that the ingredients appearing in the cooking instructions obtained with our model, have better recall and precision scores than the ingredients in retrieved instructions User Studies In the first study, the paper compares the performance of our model against human performance, in the task of recipe generation (including ingredients and recipe instructions) Result – Method outperforms both human baseline and retrieval based systems obtaining F1 of 4908% IOU and F1 Scores are used to measure the Labelling Quality

Intersection over Union ie IOU is an evaluation metric used to measure the accuracy of an annotation on a particular task whereas F1 is the harmonic mean of Precision and Recall and gives a better measure of the incorrect annotation cases The F1 score takes False Positive and False Negative into account when measuring the quality of our annotation work The second study aims at quantifying the quality of the generated recipes (ingredients and instructions), with respect to 1: the real recipes in the dataset, and 2: the ones obtained with the retrieval baseline

  Result – the success rate of generated recipes is higher than the success rate of retrieved recipes, stressing the benefits of our approach wrt retrieval Conclusion The paper introduced an image-to-recipe generation system, which takes a food image and produces a recipe consisting of a title, ingredients and sequence of cooking instructions

It first predicted sets of ingredients from food images, showing that modeling dependencies matters Then, explored instruction generation conditioned on images and inferred ingredients, highlighting the importance of reasoning about both modalities at the same time Finally, user study results confirm the difficulty of the task, and demonstrate the superiority of the system against state-of-the-art image-to-recipe retrieval approaches Thanks for watching this video The link to the paper and code can be found at the description section

Don't forget to subscribe this channel to see new videos Bye

Source: Youtube

This div height required for enabling the sticky sidebar
New Cookery Recipes
Ad Clicks : Ad Views : Ad Clicks : Ad Views :

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.

Close