# We assume you have IPFS in your system, if not plese download it first!
# start ipfs daemon to connect to the network
ipfs daemon
# run api for connecting ai and website
cd frontend\api
python save_image.py
# To deploy the website
cd frontend\chan_frontend
npm install
npm run devYou don't need to deploy the smart contract!The World NFT, aims to provide everyone with the opportunity to become a photographer and allow users worldwide to explore, purchase and use their favorite image as an NFT to own it. At the core of the platform is the use of advanced artificial intelligence technology, as well as blockchain-based smart contracts, to ensure that interactions and value exchanges between photographers and buyers are both efficient and secure. More than a simple online photo marketplace, the project represents a new, decentralized art exchange and business model.
If you are willing to upload your photography work, you will receive a modest token of appreciation as a reward. We believe that heartfelt photographic works will surely be discovered by those who love the world! We encourage everyone to join our community!
The World is here because every photoshot has a story eager for an audience. We want to give every aspiring photographer a chance to shine and every art lover a place to find that perfect image that speaks to them. By turning photos into NFTs, we're not just trading pictures; we're sharing pieces of the world. And with every photo uploaded, a token of thanks is given, encouraging more stories to be told. This is about making sure no view of the world goes unseen, and every creator feels valued. Let's make the art of photography accessible to all and celebrate the world through our lenses. Join in and let's create a gallery without walls together!
- Role: Write smart contracts primarily designed for facilitating transactions between buyers and photographers for the purchase of photographs, and capable of connecting with an oracle to transmit the photographs to an AI model.
- Role: Developing a website and researching how to enable direct interaction between our website, "The World," and smart contracts. Aim to seamlessly integrate the functionality of smart contracts into our website, allowing them to run directly on the platform.
- Role: Focused on the AI part, using CNN and other classifiers to determine if the object matches what it claims to be and also determine the quality of the picture.
- The blockchain architecture provides a decentralized platform for photographers to tokenize their work as NFTs. Sellers upload their photos along with descriptions and wallet addresses. The platform employs a machine learning model to verify that the images meet quality and description standards, ensuring authenticity. Approved photos are tokenized through a smart contract, rewarding the seller and listing the NFT for sale. Buyers can then purchase NFTs with transparency, as transactions are securely facilitated and recorded on the blockchain. Upon sale, revenue is distributed, with a percentage going to the seller.
- Smart contract is needed: Platform Contract facilitates transactions by transferring image links and prices between photographers and buyers. What's more, it can send token and recieve token to or from seller and buyer wallet addresses. Additionally, it ensures transparency and efficiency by verifying photo quality and returning revenue to the seller accordingly.
- Introducing a revolutionary online platform tailored for seamless transactions between photographers and buyers, integrating effortlessly with buyers' Metamask accounts. Our innovative system incorporates advanced AI technology to streamline the process, automatically extracting photo links and conducting evaluations in real-time. Through this streamlined interface, buyers can swiftly identify and purchase high-quality images that meet their specific needs. Additionally, the platform receives feedback from the AI model to ensure that selected photos meet the buyer's criteria in terms of quality and suitability for their intended use.
This model leverages a modified ResNet50 backbone, optimized for both classification and bounding box regression tasks in object detection:
- Backbone Network: ResNet50 without the last two layers, ending at the global average pooling layer.
- Classification Layer: A fully connected layer (
fc1) withnum_classes + 1output neurons. This includes neurons for each object class and an additional neuron for the "nothing" class. - Regression Layer: A fully connected layer (
fc2) withnum_classes * 4output neurons. Each class has four neurons representing the coordinates of bounding boxes (upper left and lower right corners).
- Data Handling: Custom datasets and
DataLoaderfor managing training, validation, and test datasets. - Preprocessing: Data augmentation and preprocessing steps include resizing and normalization, implemented via
torchvisiontransforms.
- Optimizer: Adam with a learning rate of
1e-3and weight decay of1e-5. - Loss Functions:
- Cross-entropy loss for classification.
- Mean Squared Error (MSE) loss for bounding box regression.
sequenceDiagram
actor S as Seller
participant W as Website
participant contract as Smart Contract
participant ML as ML model
actor B as Buyer
S->>+W: Upload photo
S->>contract: Set price
W->>+ML: Extract features
alt Photo quality is good
ML->>W: Return true
W->>W: Display photo
W->>S: Give a little reward
else Photo quality is bad
ML->>-W: Return false
W->>-S: Reject photo
end
B->>W: Buy photo
contract->>contract: trade approved
contract->>B: Transfer NFT
contract->>S: Transfer 60% of revenue