Image Classifier Going forward, AI algorithms will be incorporated into more and more everyday applications. For example, you might want to include an image classifier in a smartphone app. To do this, you'd use a deep learning model trained on hundreds of thousands of images as part of the overall application architecture. A large part of software development in the future will be using these types of models as common parts of applications. In this project, you'll train an image classifier to recognize different species of flowers. You can imagine using something like this in a phone app that tells you the name of the flower your camera is looking at. In practice, you'd train this classifier, then export it for use in your application. We'll be using this dataset of 102 flower categories. When you've completed this project, you'll have an application that can be trained on any set of labelled images. Here your network will be learning about flowers and end up as a command line application. But, what you do with your new skills depends on your imagination and effort in building a dataset. This is the final Project of the Udacity AI with Python Nanodegree Prerequisites The Code is written in Python 3.6.5 . If you don't have Python installed you can find it here. If you are using a lower version of Python you can upgrade using the pip package, ensuring you have the latest version of pip. To install pip run in the command Line python -m ensurepip -- default-pip to upgrade it python -m pip install -- upgrade pip setuptools wheel to upgrade Python pip install python -- upgrade Additional Packages that are required are: Numpy, Pandas, MatplotLib, Pytorch, PIL and json. You can donwload them using pip pip install numpy pandas matplotlib pil or conda conda install numpy pandas matplotlib pil In order to intall Pytorch head over to the Pytorch site select your specs and follow the instructions given. Viewing the Jyputer Notebook In order to better view and work on the jupyter Notebook I encourage you to use nbviewer . You can simply copy and paste the link to this website and you will be able to edit it without any problem. Alternatively you can clone the repository using git clone https://github.com/fotisk07/Image-Classifier/ then in the command Line type, after you have downloaded jupyter notebook type jupyter notebook locate the notebook and run it. Command Line Application Train a new network on a data set with train.py Basic Usage : python train.py data_directory Prints out current epoch, training loss, validation loss, and validation accuracy as the netowrk trains Options: Set direcotry to save checkpoints: python train.py data_dor --save_dir save_directory Choose arcitecture (alexnet, densenet121 or vgg16 available): pytnon train.py data_dir --arch "vgg16" Set hyperparameters: python train.py data_dir --learning_rate 0.001 --hidden_layer1 120 --epochs 20 Use GPU for training: python train.py data_dir --gpu gpu Predict flower name from an image with predict.py along with the probability of that name. That is you'll pass in a single image /path/to/image and return the flower name and class probability Basic usage: python predict.py /path/to/image checkpoint Options: Return top K most likely classes: python predict.py input checkpoint ---top_k 3 Use a mapping of categories to real names: python predict.py input checkpoint --category_names cat_To_name.json Use GPU for inference: python predict.py input checkpoint --gpu Json file In order for the network to print out the name of the flower a .json file is required. If you aren't familiar with json you can find information here. By using a .json file the data can be sorted into folders with numbers and those numbers will correspond to specific names specified in the .json file. Data and the json file The data used specifically for this assignemnt are a flower database are not provided in the repository as it's larger than what github allows. Nevertheless, feel free to create your own databases and train the model on them to use with your own projects. The structure of your data should be the following: The data need to comprised of 3 folders, test, train and validate. Generally the proportions should be 70% training 10% validate and 20% test. Inside the train, test and validate folders there should be folders bearing a specific number which corresponds to a specific category, clarified in the json file. For example if we have the image a.jpj and it is a rose it could be in a path like this /test/5/a.jpg and json file would be like this {...5:"rose",...}. Make sure to include a lot of photos of your catagories (more than 10) with different angles and different lighting conditions in order for the network to generalize better. GPU As the network makes use of a sophisticated deep convolutional neural network the training process is impossible to be done by a common laptop. In order to train your models to your local machine you have three options Cuda -- If you have an NVIDIA GPU then you can install CUDA from here. With Cuda you will be able to train your model however the process will still be time consuming Cloud Services -- There are many paid cloud services that let you train your models like AWS or Google Cloud Coogle Colab -- Google Colab gives you free access to a tesla K80 GPU for 12 hours at a time. Once 12 hours have ellapsed you can just reload and continue! The only limitation is that you have to upload the data to Google Drive and if the dataset is massive you may run out of space. However, once a model is trained then a normal CPU can be used for the predict.py file and you will have an answer within some seconds. Hyperparameters As you can see you have a wide selection of hyperparameters available and you can get even more by making small modifications to the code. Thus it may seem overly complicated to choose the right ones especially if the training needs at least 15 minutes to be completed. So here are some hints: By increasing the number of epochs the accuracy of the network on the training set gets better and better however be careful because if you pick a large number of epochs the network won't generalize well, that is to say it will have high accuracy on the training image and low accuracy on the test images. Eg: training for 12 epochs training accuracy: 85% Test accuracy: 82%. Training for 30 epochs training accuracy 95% test accuracy 50%. A big learning rate guarantees that the network will converge fast to a small error but it will constantly overshot A small learning rate guarantees that the network will reach greater accuracies but the learning process will take longer Densenet121 works best for images but the training process takes significantly longer than alexnet or vgg16 *My settings were lr=0.001, dropoup=0.5, epochs= 15 and my test accuracy was 86% with densenet121 as my feature extraction model. Pre-Trained Network The checkpoint.pth file contains the information of a network trained to recognise 102 different species of flowers. I has been trained with specific hyperparameters thus if you don't set them right the network will fail. In order to have a prediction for an image located in the path /path/to/image using my pretrained model you can simply type python predict.py /path/to/image checkpoint.pth Contributing Please read CONTRIBUTING.md for the process for submitting pull requests. Authors Shanmukha Mudigonda - Initial work Udacity - Final Project of the AI with Python Nanodegree

Some Jupyter notebook examples for SymPy

BUILD YOUR OWN BLOCKCHAIN: A PYTHON TUTORIAL Download the full Jupyter/iPython notebook from Github here Build Your Own Blockchain – The Basics¶ This tutorial will walk you through the basics of how to build a blockchain from scratch. Focusing on the details of a concrete example will provide a deeper understanding of the strengths and limitations of blockchains. For a higher-level overview, I’d recommend this excellent article from BitsOnBlocks. Transactions, Validation, and updating system state¶ At its core, a blockchain is a distributed database with a set of rules for verifying new additions to the database. We’ll start off by tracking the accounts of two imaginary people: Alice and Bob, who will trade virtual money with each other. We’ll need to create a transaction pool of incoming transactions, validate those transactions, and make them into a block. We’ll be using a hash function to create a ‘fingerprint’ for each of our transactions- this hash function links each of our blocks to each other. To make this easier to use, we’ll define a helper function to wrap the python hash function that we’re using. In [1]: import hashlib, json, sys def hashMe(msg=""): # For convenience, this is a helper function that wraps our hashing algorithm if type(msg)!=str: msg = json.dumps(msg,sort_keys=True) # If we don't sort keys, we can't guarantee repeatability! if sys.version_info.major == 2: return unicode(hashlib.sha256(msg).hexdigest(),'utf-8') else: return hashlib.sha256(str(msg).encode('utf-8')).hexdigest() Next, we want to create a function to generate exchanges between Alice and Bob. We’ll indicate withdrawals with negative numbers, and deposits with positive numbers. We’ll construct our transactions to always be between the two users of our system, and make sure that the deposit is the same magnitude as the withdrawal- i.e. that we’re neither creating nor destroying money. In [2]: import random random.seed(0) def makeTransaction(maxValue=3): # This will create valid transactions in the range of (1,maxValue) sign = int(random.getrandbits(1))*2 - 1 # This will randomly choose -1 or 1 amount = random.randint(1,maxValue) alicePays = sign * amount bobPays = -1 * alicePays # By construction, this will always return transactions that respect the conservation of tokens. # However, note that we have not done anything to check whether these overdraft an account return {u'Alice':alicePays,u'Bob':bobPays} Now let’s create a large set of transactions, then chunk them into blocks. In [3]: txnBuffer = [makeTransaction() for i in range(30)] Next step: making our very own blocks! We’ll take the first k transactions from the transaction buffer, and turn them into a block. Before we do that, we need to define a method for checking the valididty of the transactions we’ve pulled into the block. For bitcoin, the validation function checks that the input values are valid unspent transaction outputs (UTXOs), that the outputs of the transaction are no greater than the input, and that the keys used for the signatures are valid. In Ethereum, the validation function checks that the smart contracts were faithfully executed and respect gas limits. No worries, though- we don’t have to build a system that complicated. We’ll define our own, very simple set of rules which make sense for a basic token system: The sum of deposits and withdrawals must be 0 (tokens are neither created nor destroyed) A user’s account must have sufficient funds to cover any withdrawals If either of these conditions are violated, we’ll reject the transaction. In [4]: def updateState(txn, state): # Inputs: txn, state: dictionaries keyed with account names, holding numeric values for transfer amount (txn) or account balance (state) # Returns: Updated state, with additional users added to state if necessary # NOTE: This does not not validate the transaction- just updates the state! # If the transaction is valid, then update the state state = state.copy() # As dictionaries are mutable, let's avoid any confusion by creating a working copy of the data. for key in txn: if key in state.keys(): state[key] += txn[key] else: state[key] = txn[key] return state In [5]: def isValidTxn(txn,state): # Assume that the transaction is a dictionary keyed by account names # Check that the sum of the deposits and withdrawals is 0 if sum(txn.values()) is not 0: return False # Check that the transaction does not cause an overdraft for key in txn.keys(): if key in state.keys(): acctBalance = state[key] else: acctBalance = 0 if (acctBalance + txn[key]) < 0: return False return True Here are a set of sample transactions, some of which are fraudulent- but we can now check their validity! In [6]: state = {u'Alice':5,u'Bob':5} print(isValidTxn({u'Alice': -3, u'Bob': 3},state)) # Basic transaction- this works great! print(isValidTxn({u'Alice': -4, u'Bob': 3},state)) # But we can't create or destroy tokens! print(isValidTxn({u'Alice': -6, u'Bob': 6},state)) # We also can't overdraft our account. print(isValidTxn({u'Alice': -4, u'Bob': 2,'Lisa':2},state)) # Creating new users is valid print(isValidTxn({u'Alice': -4, u'Bob': 3,'Lisa':2},state)) # But the same rules still apply! True False False True False Each block contains a batch of transactions, a reference to the hash of the previous block (if block number is greater than 1), and a hash of its contents and the header Building the Blockchain: From Transactions to Blocks¶ We’re ready to start making our blockchain! Right now, there’s nothing on the blockchain, but we can get things started by defining the ‘genesis block’ (the first block in the system). Because the genesis block isn’t linked to any prior block, it gets treated a bit differently, and we can arbitrarily set the system state. In our case, we’ll create accounts for our two users (Alice and Bob) and give them 50 coins each. In [7]: state = {u'Alice':50, u'Bob':50} # Define the initial state genesisBlockTxns = [state] genesisBlockContents = {u'blockNumber':0,u'parentHash':None,u'txnCount':1,u'txns':genesisBlockTxns} genesisHash = hashMe( genesisBlockContents ) genesisBlock = {u'hash':genesisHash,u'contents':genesisBlockContents} genesisBlockStr = json.dumps(genesisBlock, sort_keys=True) Great! This becomes the first element from which everything else will be linked. In [8]: chain = [genesisBlock] For each block, we want to collect a set of transactions, create a header, hash it, and add it to the chain In [9]: def makeBlock(txns,chain): parentBlock = chain[-1] parentHash = parentBlock[u'hash'] blockNumber = parentBlock[u'contents'][u'blockNumber'] + 1 txnCount = len(txns) blockContents = {u'blockNumber':blockNumber,u'parentHash':parentHash, u'txnCount':len(txns),'txns':txns} blockHash = hashMe( blockContents ) block = {u'hash':blockHash,u'contents':blockContents} return block Let’s use this to process our transaction buffer into a set of blocks: In [10]: blockSizeLimit = 5 # Arbitrary number of transactions per block- # this is chosen by the block miner, and can vary between blocks! while len(txnBuffer) > 0: bufferStartSize = len(txnBuffer) ## Gather a set of valid transactions for inclusion txnList = [] while (len(txnBuffer) > 0) & (len(txnList) < blockSizeLimit): newTxn = txnBuffer.pop() validTxn = isValidTxn(newTxn,state) # This will return False if txn is invalid if validTxn: # If we got a valid state, not 'False' txnList.append(newTxn) state = updateState(newTxn,state) else: print("ignored transaction") sys.stdout.flush() continue # This was an invalid transaction; ignore it and move on ## Make a block myBlock = makeBlock(txnList,chain) chain.append(myBlock) In [11]: chain[0] Out[11]: {'contents': {'blockNumber': 0, 'parentHash': None, 'txnCount': 1, 'txns': [{'Alice': 50, 'Bob': 50}]}, 'hash': '7c88a4312054f89a2b73b04989cd9b9e1ae437e1048f89fbb4e18a08479de507'} In [12]: chain[1] Out[12]: {'contents': {'blockNumber': 1, 'parentHash': '7c88a4312054f89a2b73b04989cd9b9e1ae437e1048f89fbb4e18a08479de507', 'txnCount': 5, 'txns': [{'Alice': 3, 'Bob': -3}, {'Alice': -1, 'Bob': 1}, {'Alice': 3, 'Bob': -3}, {'Alice': -2, 'Bob': 2}, {'Alice': 3, 'Bob': -3}]}, 'hash': '7a91fc8206c5351293fd11200b33b7192e87fad6545504068a51aba868bc6f72'} As expected, the genesis block includes an invalid transaction which initiates account balances (creating tokens out of thin air). The hash of the parent block is referenced in the child block, which contains a set of new transactions which affect system state. We can now see the state of the system, updated to include the transactions: In [13]: state Out[13]: {'Alice': 72, 'Bob': 28} Checking Chain Validity¶ Now that we know how to create new blocks and link them together into a chain, let’s define functions to check that new blocks are valid- and that the whole chain is valid. On a blockchain network, this becomes important in two ways: When we initially set up our node, we will download the full blockchain history. After downloading the chain, we would need to run through the blockchain to compute the state of the system. To protect against somebody inserting invalid transactions in the initial chain, we need to check the validity of the entire chain in this initial download. Once our node is synced with the network (has an up-to-date copy of the blockchain and a representation of system state) it will need to check the validity of new blocks that are broadcast to the network. We will need three functions to facilitate in this: checkBlockHash: A simple helper function that makes sure that the block contents match the hash checkBlockValidity: Checks the validity of a block, given its parent and the current system state. We want this to return the updated state if the block is valid, and raise an error otherwise. checkChain: Check the validity of the entire chain, and compute the system state beginning at the genesis block. This will return the system state if the chain is valid, and raise an error otherwise. In [14]: def checkBlockHash(block): # Raise an exception if the hash does not match the block contents expectedHash = hashMe( block['contents'] ) if block['hash']!=expectedHash: raise Exception('Hash does not match contents of block %s'% block['contents']['blockNumber']) return In [15]: def checkBlockValidity(block,parent,state): # We want to check the following conditions: # - Each of the transactions are valid updates to the system state # - Block hash is valid for the block contents # - Block number increments the parent block number by 1 # - Accurately references the parent block's hash parentNumber = parent['contents']['blockNumber'] parentHash = parent['hash'] blockNumber = block['contents']['blockNumber'] # Check transaction validity; throw an error if an invalid transaction was found. for txn in block['contents']['txns']: if isValidTxn(txn,state): state = updateState(txn,state) else: raise Exception('Invalid transaction in block %s: %s'%(blockNumber,txn)) checkBlockHash(block) # Check hash integrity; raises error if inaccurate if blockNumber!=(parentNumber+1): raise Exception('Hash does not match contents of block %s'%blockNumber) if block['contents']['parentHash'] != parentHash: raise Exception('Parent hash not accurate at block %s'%blockNumber) return state In [16]: def checkChain(chain): # Work through the chain from the genesis block (which gets special treatment), # checking that all transactions are internally valid, # that the transactions do not cause an overdraft, # and that the blocks are linked by their hashes. # This returns the state as a dictionary of accounts and balances, # or returns False if an error was detected ## Data input processing: Make sure that our chain is a list of dicts if type(chain)==str: try: chain = json.loads(chain) assert( type(chain)==list) except: # This is a catch-all, admittedly crude return False elif type(chain)!=list: return False state = {} ## Prime the pump by checking the genesis block # We want to check the following conditions: # - Each of the transactions are valid updates to the system state # - Block hash is valid for the block contents for txn in chain[0]['contents']['txns']: state = updateState(txn,state) checkBlockHash(chain[0]) parent = chain[0] ## Checking subsequent blocks: These additionally need to check # - the reference to the parent block's hash # - the validity of the block number for block in chain[1:]: state = checkBlockValidity(block,parent,state) parent = block return state We can now check the validity of the state: In [17]: checkChain(chain) Out[17]: {'Alice': 72, 'Bob': 28} And even if we are loading the chain from a text file, e.g. from backup or loading it for the first time, we can check the integrity of the chain and create the current state: In [18]: chainAsText = json.dumps(chain,sort_keys=True) checkChain(chainAsText) Out[18]: {'Alice': 72, 'Bob': 28} Putting it together: The final Blockchain Architecture¶ In an actual blockchain network, new nodes would download a copy of the blockchain and verify it (as we just did above), then announce their presence on the peer-to-peer network and start listening for transactions. Bundling transactions into a block, they then pass their proposed block on to other nodes. We’ve seen how to verify a copy of the blockchain, and how to bundle transactions into a block. If we recieve a block from somewhere else, verifying it and adding it to our blockchain is easy. Let’s say that the following code runs on Node A, which mines the block: In [19]: import copy nodeBchain = copy.copy(chain) nodeBtxns = [makeTransaction() for i in range(5)] newBlock = makeBlock(nodeBtxns,nodeBchain) Now assume that the newBlock is transmitted to our node, and we want to check it and update our state if it is a valid block: In [20]: print("Blockchain on Node A is currently %s blocks long"%len(chain)) try: print("New Block Received; checking validity...") state = checkBlockValidity(newBlock,chain[-1],state) # Update the state- this will throw an error if the block is invalid! chain.append(newBlock) except: print("Invalid block; ignoring and waiting for the next block...") print("Blockchain on Node A is now %s blocks long"%len(chain)) Blockchain on Node A is currently 7 blocks long New Block Received; checking validity... Blockchain on Node A is now 8 blocks long

Some Jupyter Notebooks I am creating for the course Applied Elasticity at IIT Kharagpur