Question 280
A research agent must independently analyze pricing, support tickets, and customer sentiment before combining findings. What orchestration mode is suitable?
- Use sequential orchestration with explicit dependencies between each stage
- Run independent analysis branches in parallel and join their results
- Capture step-level traces, tool inputs, outputs, and terminal outcomes
- Use a routing or triage step based on the user's intent and allowed capabilities
Explanation
Correct Answer
B:Independent tasks can execute concurrently before a synchronization step produces the combined report. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after running independent analysis branches in parallel and join their results.
Incorrect Answers
A:That change addresses another failure mode: downstream work depends on earlier validated outputs, so parallel execution risks using incomplete data; it does not address this case, for which the relevant intervention is to run independent analysis branches in parallel and join their results.
C:Although useful elsewhere in the multi-step operations assistant, workflow correctness depends on actions taken during the run, not just the fluent final message; the present case is different and calls for the team to run independent analysis branches in parallel and join their results.
D:Triage maps a request to the responsible worker rather than running every specialist indiscriminately in its own workflow, but that is a different correction from the present requirement to run independent analysis branches in parallel and join their results. The reported issue would remain.
Question 281
A model returns one large box around stacked products while workers need each item individually. Which approach may be required?
- Request or calculate separate focal crops for each required aspect ratio
- Train a custom vision classification model using labeled domain images
- Test caption quality on domain-specific images and add explicit OCR or detection for critical details
- Evaluate an instance-capable object detector trained on representative stacked-product images
Explanation
Correct Answer
D:Generic scene tags or coarse detection may not separate densely packed custom objects. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after evaluating an instance-capable object detector trained on representative stacked-product images.
Incorrect Answers
A:That change addresses another failure mode: one crop rectangle cannot necessarily preserve important content across very different display shapes; it does not address this case, for which the relevant intervention is to evaluate an instance-capable object detector trained on representative stacked-product images.
B:Although useful elsewhere in the e-commerce image processing service, generic tags use a predefined vocabulary; supervised custom classes are needed for specialized taxonomy; the present case is different and calls for the team to evaluate an instance-capable object detector trained on representative stacked-product images.
C:General captions optimize overall visual description and can miss small business-critical marks in its own workflow, but that is a different correction from the present requirement to evaluate an instance-capable object detector trained on representative stacked-product images. The reported issue would remain.
Question 282
Users report that generated captions omit important small symbols. What evaluation step helps?
- Evaluate an instance-capable object detector trained on representative stacked-product images
- Improve resolution or reacquire the image and validate low-confidence text
- Request or calculate separate focal crops for each required aspect ratio
- Test caption quality on domain-specific images and add explicit OCR or detection for critical details
Explanation
Correct Answer
D:General captions optimize overall visual description and can miss small business-critical marks. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after testing caption quality on domain-specific images and add explicit OCR or detection for critical details.
Incorrect Answers
A:Generic scene tags or coarse detection may not separate densely packed custom objects in its own workflow, but that is a different correction from the present requirement to test caption quality on domain-specific images and add explicit OCR or detection for critical details. The reported issue would remain.
B:Although useful elsewhere in the e-commerce image processing service, poor image quality reduces text recognition accuracy; a different classification threshold will not reconstruct missing pixels; the present case is different and calls for the team to test caption quality on domain-specific images and add explicit OCR or detection for critical details.
C:That change addresses another failure mode: one crop rectangle cannot necessarily preserve important content across very different display shapes; it does not address this case, for which the relevant intervention is to test caption quality on domain-specific images and add explicit OCR or detection for critical details.
Question 283
A suggested crop excludes a product label the business must show. What should the pipeline do?
- Transform returned bounding coordinates into the displayed image dimensions correctly
- Do not treat image captions as reliable identity verification
- Respect the returned reading-order and layout metadata where available
- Apply business-specific region checks or override the suggested crop when needed
Explanation
Correct Answer
D:An aesthetically prominent region does not guarantee inclusion of regulatory or SKU labels. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after applying business-specific region checks or override the suggested crop when needed.
Incorrect Answers
A:Coordinate geometry must be mapped from the inference image frame to the rendered viewport in its own workflow, but that is a different correction from the present requirement to apply business-specific region checks or override the suggested crop when needed. The reported issue would remain.
B:Although useful elsewhere in the e-commerce image processing service, visual description is not authorization or a verified identity match; the present case is different and calls for the team to apply business-specific region checks or override the suggested crop when needed.
C:That change addresses another failure mode: extracted text geometry matters for reconstructing multi-column or rotated content; it does not address this case, for which the relevant intervention is to apply business-specific region checks or override the suggested crop when needed.
Question 284
The business requires its own fine-grained labels not present in generic tags. What should it investigate?
- Train a custom vision classification model using labeled domain images
- Request or calculate separate focal crops for each required aspect ratio
- Improve resolution or reacquire the image and validate low-confidence text
- Evaluate an instance-capable object detector trained on representative stacked-product images
Explanation
Correct Answer
A:Generic tags use a predefined vocabulary; supervised custom classes are needed for specialized taxonomy. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after training a custom vision classification model using labeled domain images.
Incorrect Answers
B:Although useful elsewhere in the e-commerce image processing service, one crop rectangle cannot necessarily preserve important content across very different display shapes; the present case is different and calls for the team to train a custom vision classification model using labeled domain images.
C:Poor image quality reduces text recognition accuracy; a different classification threshold will not reconstruct missing pixels in its own workflow, but that is a different correction from the present requirement to train a custom vision classification model using labeled domain images. The reported issue would remain.
D:That change addresses another failure mode: generic scene tags or coarse detection may not separate densely packed custom objects; it does not address this case, for which the relevant intervention is to train a custom vision classification model using labeled domain images.
Question 285
A camera image is blurry and OCR produces uncertain serial numbers. What should the capture workflow do?
- Test caption quality on domain-specific images and add explicit OCR or detection for critical details
- Train a custom vision classification model using labeled domain images
- Evaluate an instance-capable object detector trained on representative stacked-product images
- Improve resolution or reacquire the image and validate low-confidence text
Explanation
Correct Answer
D:Poor image quality reduces text recognition accuracy; a different classification threshold will not reconstruct missing pixels. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after improving resolution or reacquire the image and validate low-confidence text.
Incorrect Answers
A:That change addresses another failure mode: general captions optimize overall visual description and can miss small business-critical marks; it does not address this case, for which the relevant intervention is to improve resolution or reacquire the image and validate low-confidence text.
B:Generic tags use a predefined vocabulary; supervised custom classes are needed for specialized taxonomy in its own workflow, but that is a different correction from the present requirement to improve resolution or reacquire the image and validate low-confidence text. The reported issue would remain.
C:Although useful elsewhere in the e-commerce image processing service, generic scene tags or coarse detection may not separate densely packed custom objects; the present case is different and calls for the team to improve resolution or reacquire the image and validate low-confidence text.
Question 286
An e-commerce image processing service must locate instances of supported object classes in the image. Reviewers will accept the implementation when each recognized object has associated location information. Which approach best fulfills that requirement?
- Request image captioning
- Enable Image Analysis OCR or Read
- Request image tags
- Request object detection
- Use smart crop suggestions
Explanation
Correct Answer
D:Object detection returns object labels with spatial regions or bounding boxes when those categories are supported. For this e-commerce image processing service, the intended evidence is that each recognized object has associated location information. It does not simply summarize the whole photograph in fluent prose.
Incorrect Answers
A:Captioning produces a descriptive text summary of image content using supported Image Analysis capabilities; it may be used to generate a natural-language sentence describing the overall visual scene in a e-commerce image processing service, but the current requirement is instead to locate instances of supported object classes in the image. That separate capability does not prove that each recognized object has associated location information; It does not provide guaranteed verbatim OCR for small serial numbers, so selecting that alternative would not meet the stated acceptance check.
B:OCR recognizes textual content in images and returns recognized text with spatial information where supported; it may be used to extract the exact words printed or handwritten in a product photograph in a e-commerce image processing service, but the current requirement is instead to locate instances of supported object classes in the image. That separate capability does not prove that each recognized object has associated location information; It does not inherently return product taxonomies or detect independent objects by label, so selecting that alternative would not meet the stated acceptance check.
C:Image tagging supplies concise labels corresponding to concepts or objects visible in an image; it may be used to obtain broad descriptive categories associated with the picture in a e-commerce image processing service, but the current requirement is instead to locate instances of supported object classes in the image. That separate capability does not prove that each recognized object has associated location information; Tags are not the same as a prose caption or object-by-object bounding boxes, so selecting that alternative would not meet the stated acceptance check.
E:Smart crop estimates useful image regions for requested aspect ratios so thumbnails focus on salient content; it may be used to identify an aspect-ratio-aware crop that retains important visual content in a e-commerce image processing service, but the current requirement is instead to locate instances of supported object classes in the image. That separate capability does not prove that each recognized object has associated location information; It does not perform barcode recognition or produce semantic embeddings by itself, so selecting that alternative would not meet the stated acceptance check.
Question 287
A retailer needs descriptions of several distinct regions within a busy shelf image, not one global scene summary. What supported feature should it evaluate?
- Use OCR or Read text extraction and consume recognized lines and words
- Consume separate detected object instances with their bounding regions
- Request dense image captions or another region-aware description capability when available
- Request supported image tagging and store the returned tag labels
Explanation
Correct Answer
C:A single global caption can omit local items; dense or region-level descriptions convey additional spatially localized visual content. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after requesting dense image captions or another region-aware description capability when available.
Incorrect Answers
A:Scene captioning summarizes an image; OCR targets the exact characters on printed labels in its own workflow, but that is a different correction from the present requirement to request dense image captions or another region-aware description capability when available. The reported issue would remain.
B:That change addresses another failure mode: per-object detections support counting and localization; one image-level class does not distinguish instances; it does not address this case, for which the relevant intervention is to request dense image captions or another region-aware description capability when available.
D:Although useful elsewhere in the e-commerce image processing service, tags summarize visual concepts, while object detection targets located items with regions; the present case is different and calls for the team to request dense image captions or another region-aware description capability when available.
Question 288
Object detections are returned in the source-image coordinate frame, but a web UI displays the image with resizing and letterbox padding. Which TWO transformations must be handled? (Choose TWO.)
- Apply business-specific region checks or override the suggested crop when needed
- Scale detection coordinates into the displayed image dimensions
- Do not treat image captions as reliable identity verification
- Respect the returned reading-order and layout metadata where available
- Adjust bounding boxes for the padding introduced by letterboxing
Explanation
Correct Answers
B:Coordinates tied to the original image size must be transformed proportionally before drawing correct UI overlays.
E:The display includes offset margins beyond the scaled image, so scaling without padding correction shifts every detection rectangle.
Incorrect Answers
A:An aesthetically prominent region does not guarantee inclusion of regulatory or SKU labels in its own workflow, but that is a different correction from the present requirement to scale detection coordinates into the displayed image dimensions and adjust bounding boxes for the padding introduced by letterboxing. The reported issue would remain.
C:That change addresses another failure mode: visual description is not authorization or a verified identity match; it does not address this case, for which the relevant intervention is to scale detection coordinates into the displayed image dimensions and adjust bounding boxes for the padding introduced by letterboxing.
D:Although useful elsewhere in the e-commerce image processing service, extracted text geometry matters for reconstructing multi-column or rotated content; the present case is different and calls for the team to scale detection coordinates into the displayed image dimensions and adjust bounding boxes for the padding introduced by letterboxing.
Question 289
A shop catalog needs broad descriptors such as “footwear” and “outdoors,” not a list of rectangles. What visual output should be used?
- Request dense image captions or another region-aware description capability when available
- Consume separate detected object instances with their bounding regions
- Request supported image tagging and store the returned tag labels
- Use OCR or Read text extraction and consume recognized lines and words
Explanation
Correct Answer
C:Tags summarize visual concepts, while object detection targets located items with regions. In the e-commerce image processing service, confirm the fix by retesting the affected behavior after requesting supported image tagging and store the returned tag labels.
Incorrect Answers
A:Although useful elsewhere in the e-commerce image processing service, a single global caption can omit local items; dense or region-level descriptions convey additional spatially localized visual content; the present case is different and calls for the team to request supported image tagging and store the returned tag labels.
B:That change addresses another failure mode: per-object detections support counting and localization; one image-level class does not distinguish instances; it does not address this case, for which the relevant intervention is to request supported image tagging and store the returned tag labels.
D:Scene captioning summarizes an image; OCR targets the exact characters on printed labels in its own workflow, but that is a different correction from the present requirement to request supported image tagging and store the returned tag labels. The reported issue would remain.
Question 290
A release for an e-commerce image processing service is blocked until the application can extract the exact words printed or handwritten in a product photograph. The tester will verify that the output contains recognized text segments and positions. What should be implemented?
- Request image tags
- Enable Image Analysis OCR or Read
- Request image captioning
- Request object detection
Explanation
Correct Answer
B:OCR recognizes textual content in images and returns recognized text with spatial information where supported. For this e-commerce image processing service, the intended evidence is that the output contains recognized text segments and positions. It does not inherently return product taxonomies or detect independent objects by label.
Incorrect Answers
A:Image tagging supplies concise labels corresponding to concepts or objects visible in an image; it may be used to obtain broad descriptive categories associated with the picture in a e-commerce image processing service, but the current requirement is instead to extract the exact words printed or handwritten in a product photograph. That separate capability does not prove that the output contains recognized text segments and positions; Tags are not the same as a prose caption or object-by-object bounding boxes, so selecting that alternative would not meet the stated acceptance check.
C:Captioning produces a descriptive text summary of image content using supported Image Analysis capabilities; it may be used to generate a natural-language sentence describing the overall visual scene in a e-commerce image processing service, but the current requirement is instead to extract the exact words printed or handwritten in a product photograph. That separate capability does not prove that the output contains recognized text segments and positions; It does not provide guaranteed verbatim OCR for small serial numbers, so selecting that alternative would not meet the stated acceptance check.
D:Object detection returns object labels with spatial regions or bounding boxes when those categories are supported; it may be used to locate instances of supported object classes in the image in a e-commerce image processing service, but the current requirement is instead to extract the exact words printed or handwritten in a product photograph. That separate capability does not prove that the output contains recognized text segments and positions; It does not simply summarize the whole photograph in fluent prose, so selecting that alternative would not meet the stated acceptance check.