Uncaught Error: The dict provided in model.execute(dict) has keys that are not part of graph

model used: custom training ultralytics yolo v5 converted to tensorflow saved model then to tensorflow js (quantized to uint8) when using model.predict getting this error,

• OS Platform: Windows 10 Desktop Chrome
• TensorFlow.js installed from (npm or script link): 2.7.0
• Browser version: Version 86.0.4240.183 (Official Build) (64-bit)
• Tensorflow.js Converter Version: 2.7.0

Uncaught Error: The dict provided in model.execute(dict) has keys: [alt,src,srcset,sizes,crossOrigin,useMap,isMap,width,height,naturalWidth,naturalHeight,complete,currentSrc,referrerPolicy,decoding,name,lowsrc,align,hspace,vspace,longDesc,border,x,y,decode,loading,title,lang,translate,dir,hidden,accessKey,draggable,spellcheck,autocapitalize,contentEditable,isContentEditable,inputMode,offsetParent,offsetTop,offsetLeft,offsetWidth,offsetHeight,style,innerText,outerText,onabort,onblur,oncancel,oncanplay,oncanplaythrough,onchange,onclick,onclose,oncontextmenu,oncuechange,ondblclick,ondrag,ondragend,ondragenter,ondragleave,ondragover,ondragstart,ondrop,ondurationchange,onemptied,onended,onerror,onfocus,onformdata,oninput,oninvalid,onkeydown,onkeypress,onkeyup,onload,onloadeddata,onloadedmetadata,onloadstart,onmousedown,onmouseenter,onmouseleave,onmousemove,onmouseout,onmouseover,onmouseup,onmousewheel,onpause,onplay,onplaying,onprogress,onratechange,onreset,onresize,onscroll,onseeked,onseeking,onselect,onstalled,onsubmit,onsuspend,ontimeupdate,ontoggle,onvolumechange,onwaiting,onwebkitanimationend,onwebkitanimationiteration,onwebkitanimationstart,onwebkittransitionend,onwheel,onauxclick,ongotpointercapture,onlostpointercapture,onpointerdown,onpointermove,onpointerup,onpointercancel,onpointerover,onpointerout,onpointerenter,onpointerleave,onselectstart,onselectionchange,onanimationend,onanimationiteration,onanimationstart,ontransitionend,oncopy,oncut,onpaste,dataset,nonce,autofocus,tabIndex,attachInternals,blur,click,focus,onpointerrawupdate,enterKeyHint,namespaceURI,prefix,localName,tagName,id,className,classList,slot,attributes,shadowRoot,part,assignedSlot,innerHTML,outerHTML,scrollTop,scrollLeft,scrollWidth,scrollHeight,clientTop,clientLeft,clientWidth,clientHeight,attributeStyleMap,onbeforecopy,onbeforecut,onbeforepaste,onsearch,elementTiming,onfullscreenchange,onfullscreenerror,onwebkitfullscreenchange,onwebkitfullscreenerror,onbeforexrselect,children,firstElementChild,lastElementChild,childElementCount,previousElementSibling,nextElementSibling,after,animate,append,attachShadow,before,closest,computedStyleMap,getAttribute,getAttributeNS,getAttributeNames,getAttributeNode,getAttributeNodeNS,getBoundingClientRect,getClientRects,getElementsByClassName,getElementsByTagName,getElementsByTagNameNS,hasAttribute,hasAttributeNS,hasAttributes,hasPointerCapture,insertAdjacentElement,insertAdjacentHTML,insertAdjacentText,matches,prepend,querySelector,querySelectorAll,releasePointerCapture,remove,removeAttribute,removeAttributeNS,removeAttributeNode,replaceWith,requestFullscreen,requestPointerLock,scroll,scrollBy,scrollIntoView,scrollIntoViewIfNeeded,scrollTo,setAttribute,setAttributeNS,setAttributeNode,setAttributeNodeNS,setPointerCapture,toggleAttribute,webkitMatchesSelector,webkitRequestFullScreen,webkitRequestFullscreen,ariaDescription,ariaAtomic,ariaAutoComplete,ariaBusy,ariaChecked,ariaColCount,ariaColIndex,ariaColSpan,ariaCurrent,ariaDisabled,ariaExpanded,ariaHasPopup,ariaHidden,ariaKeyShortcuts,ariaLabel,ariaLevel,ariaLive,ariaModal,ariaMultiLine,ariaMultiSelectable,ariaOrientation,ariaPlaceholder,ariaPosInSet,ariaPressed,ariaReadOnly,ariaRelevant,ariaRequired,ariaRoleDescription,ariaRowCount,ariaRowIndex,ariaRowSpan,ariaSelected,ariaSetSize,ariaSort,ariaValueMax,ariaValueMin,ariaValueNow,ariaValueText,getAnimations,replaceChildren,nodeType,nodeName,baseURI,isConnected,ownerDocument,parentNode,parentElement,childNodes,firstChild,lastChild,previousSibling,nextSibling,nodeValue,textContent,ELEMENT_NODE,ATTRIBUTE_NODE,TEXT_NODE,CDATA_SECTION_NODE,ENTITY_REFERENCE_NODE,ENTITY_NODE,PROCESSING_INSTRUCTION_NODE,COMMENT_NODE,DOCUMENT_NODE,DOCUMENT_TYPE_NODE,DOCUMENT_FRAGMENT_NODE,NOTATION_NODE,DOCUMENT_POSITION_DISCONNECTED,DOCUMENT_POSITION_PRECEDING,DOCUMENT_POSITION_FOLLOWING,DOCUMENT_POSITION_CONTAINS,DOCUMENT_POSITION_CONTAINED_BY,DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC,appendChild,cloneNode,compareDocumentPosition,contains,getRootNode,hasChildNodes,insertBefore,isDefaultNamespace,isEqualNode,isSameNode,lookupNamespaceURI,lookupPrefix,normalize,removeChild,replaceChild,addEventListener,dispatchEvent,removeEventListener] that are not part of graph at e.t.checkInputs (graph_executor.js:607) at e.t.execute (graph_executor.js:193) at e.t.execute (graph_model.js:323) at e.t.predict (graph_model.js:276) at HTMLImageElement.handleClick (index.js:100) t.checkInputs @ graph_executor.js:607 t.execute @ graph_executor.js:193 t.execute @ graph_model.js:323 t.predict @ graph_model.js:276 handleClick @ index.js:100

[patlevin]

@prakashr85 You are passing the wrong data into the model. The model expects a tensor but you pass it a DOM element.

The demo works, because it uses a custom model class that abstracts the input and accepts tensors as well as DOM elements:

The demo code (line 67):

model.detect(event.target).then(function (predictions) { …

Your code (line 63)

model.predict(event.target).then(function (predictions) { …

The difference might be subtle, but the TFJS model doesn’t have a detect-method - that’s a method implemented in the custom code for the model.

You can use the following code snipped to get the correct input for model.predict:

const input = tf.tidy(() => {
  const img = tf.browser.fromPixels(event.target)
  return img.expandDims(0)
})

model.predict(input).then(function(results) { ...

I doubt that would be all of the necessary changes, though, since the model will return an array of tensors that you will need to process according to the model’s architecture. Judging from the ultralytics yolo v5 source code, you will have to translate the relevant parts of yolo.py and test.py to JavaScript (or TypeScript) in order to use it in the browser.

[google-ml-butler[bot]]

Are you satisfied with the resolution of your issue? Yes No