tigerocr is a client used for benchmarking OCR performance of leading cloud providers.
go install ./...
Follow the documentation below to create keys for AWS, Azure, and GCP
$ tigerocr run --help
usage: tigerocr run [-keys=~/keydir/] [-aws] [-azure] [-gcp] image.jpg
-aws
Run AWS Textract OCR. Key files: credentials config
More info: https://docs.aws.amazon.com/textract/latest/dg/setup-awscli-sdk.html
-azure
Run Azure CognitiveServices OCR. Key file: azure.json
More info: https://docs.microsoft.com/azure/cognitive-services/cognitive-services-apis-create-account
Note: Create a json file with 'subscription_key' and 'endpoint' items
-azureR
Run Azure CognitiveServices Read API.
-gcp
Run GCP Vision OCR. Key file: gcp.json
More info: https://cloud.google.com/vision/docs/before-you-begin
-keys string
Path to credentials directory (default "~/.aws")
$ tigerocr --help
usage: tigerocr <command> [arguments]
The commands are:
run execute ocr on selected providers
annotate draw bounding boxes of words on the original image
editdist calculate levenshtein distance of two text files
convert convert json ocr responses to unified blw format (*)
extract extract metadata from a blw or json datafile
explore execute pdf ocr and output results as a web explorer
serve serve current directory at 127.0.0.1:8080
$ tigerocr explore -keys ~/.aws -aws -azure -azureR -gcp book.pdf
[INFO] PDF to PNG (Total: 20) ... 32 secs
[INFO] PDF to TXT (Total: 20) ... 9 secs
[ATTN] Estimate: $0.12 (80 ops). Run? [N/y]: y
[INFO] Executing OCR (Total: 80) ... 56 secs
[INFO] JSON to BLW (Total: 80) ... 5 secs
[INFO] BLW to TXT ... 0 secs
[INFO] Running Levenshtein Distance ... 1 secs
[INFO] Creating Explorer ... done
[INFO] Comparable Ptrs: 20 (out of 20). aws: 20, azu: 20, azuR: 20, gcp: 20
[DONE] Run: tigerocr serve ./explorer-book
$ tigerocr serve ./explorer-book