Rekognition: error when trying to detect faces with s3 object name containing a colon (:)

Question

Actually Rekognition works fine but when I use a filename containing a colon (:) for the S3Object, it makes an error. It is very problematic for me because all my files already have colons and I can't change their names.

So if use this It works fine:
```
{
 "Image":{
   "S3Object":{
   "Bucket":"console-sample-images",
   "Name":"skateboard.jpg"
 }
```
 but if i use a name with a colon like this It gives me an error.
```
{
 "Image":{
   "S3Object":{
   "Bucket":"console-sample-images",
   "Name":"skate:board.jpg"
 }
```
Error output:
`{"name":"Error","content":"{\"__type\":\"InvalidS3ObjectException\",\"Code\":\"InvalidS3ObjectException\",\"Message\":\"Unable to get object metadata from S3. Check object key, region and/or access permissions.\"}","message":"faultCode:Server.Error.Request faultString:'null' faultDetail:'null'","rootCause":{"errorID":2032,"target":{"bytesLoaded":174,"dataFormat":"text","bytesTotal":174,"data":"{\"__type\":\"InvalidS3ObjectException\",\"Code\":\"InvalidS3ObjectException\",\"Message\":\"Unable to get object metadata from S3. Check object key, region and/or access permissions.\"}"},"text":"Error #2032: Stream Error. URL: https://rekognition.eu-west-1.amazonaws.com","currentTarget":{"bytesLoaded":174,"dataFormat":"text","bytesTotal":174,"data":"{\"__type\":\"InvalidS3ObjectException\",\"Code\":\"InvalidS3ObjectException\",\"Message\":\"Unable to get object metadata from S3. Check object key, region and/or access permissions.\"}"},"type":"ioError","bubbles":false,"eventPhase":2,"cancelable":false},"errorID":0,"faultCode":"Server.Error.Request","faultDetail":null,"faultString":""}`

Is there a workaround for this problem? (encoding the ':' a certain way?)

Thank you for your help.

Answer

Thank you for your help.
Indeed the problem doesn't come frome Rekognition.
I don't know why but the code I used added a space after the colon. I guess it was for the  json readability.
Best

Answer

Can you update your question with the exact error/stack-trace?

*Editing answer to reflect latest update -*

```
aws rekognition detect-faces \
--region us-east-1 \
--image '{"S3Object":{"Bucket":"test-images","Name":"drive:resized.jpg"}}'
```

works as expected 
```
{
    "FaceDetails": [
        {
            "BoundingBox": {
                "Width": 0.20745284855365753,
                "Height": 0.4039246439933777,
                "Left": 0.15659691393375397,
                "Top": 0.13662810623645782
            }...
```

Can you double check S3 permissions while calling the API? Do the credentials that are being used to call the RekognitionAPI have get/read access to the S3 bucket?

~~General advice -
The object naming conventions are defined by S3 - https://docs.aws.amazon.com/AmazonS3/latest/userguide/object-keys.html~~

~~The above link suggests that url encoding special characters might help. Can you try replacing ":" in your request with "%3A"?~~

Rekognition: error when trying to detect faces with s3 object name containing a colon (:)

Relevant content